Embed Size (px)
Citation preview
NOVEL APPROACHES TO AUTOMATED QUALITY CONTROL
ANALYSES OF EDIBLE OILS BY FOURIER TRANSFORM
INFRARED SPECTROSCOPY: DETERMINATION OF FREE
FATTY ACID AND MOISTURE CONTENT
Ahmed Ali AI-Alawi
September 2005
Department of Food Science and Agricultural Chemistry
Mc Gill University, Montreal
A thesis submitted to the Faculty of Graduate Studies and Research in partial fulfillment
of the requirements of the degree of Doctor of Philosophy
©Ahmed Ali AI-Alawi, 2005
1+1 Library and Archives Canada
Bibliothèque et Archives Canada
Published Heritage Branch
Direction du Patrimoine de l'édition
395 Wellington Street Ottawa ON K1A ON4 Canada
395, rue Wellington Ottawa ON K1A ON4 Canada
NOTICE: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats.
The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
ln compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis.
While these forms may be included in the document page cou nt, their removal does not represent any loss of content from the thesis.
• •• Canada
AVIS:
Your file Votre référence ISBN: 978-0-494-25088-4 Our file Notre référence ISBN: 978-0-494-25088-4
L'auteur a accordé une licence non exclusive permettant à la Bibliothèque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par télécommunication ou par l'Internet, prêter, distribuer et vendre des thèses partout dans le monde, à des fins commerciales ou autres, sur support microforme, papier, électronique et/ou autres formats.
L'auteur conserve la propriété du droit d'auteur et des droits moraux qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Conformément à la loi canadienne sur la protection de la vie privée, quelques formulaires secondaires ont été enlevés de cette thèse.
Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant.
Short title:
FTIR ANAL YSIS OF EDIBLE OILS FOR FREE FATTY ACIDS AND MOISTURE
ABSTRACT
Three new quantitative Fourier transform infrared (FTIR) spectroscopie methods
were developed for the analysis of edible oils: two procedures to measure free fattyacids
(FFA) and one to measure moisture (HzO), the latter two methods ultimately being
automated and implemented on an auto-sampler equipped FTIR spectrometer. The
methods developed for FF A determination both convert FF As to their carboxylate salts
by me ans of acid/base reaction without causing oil saponification, one approach using 1-
propanol, an oil-miscible solvent, and the other using methanol, an oil-immiscible solvent
into which the FF A salts are extracted. The first method involves splitting oil samples
into two halves, with one half treated with propanol containing base and the other half
with propanol only. The spectra of each half is coHected and a differential spectrum
obtained, from which quantization is performed. The methanol procedure simply involves
extracting FF A into methanol containing a weak base and quantitating the FF A salts
produced. Both FF A methods determine the FF A content by measuring the v (COO")
absorbance at ~1570 cm-I relative to a reference wavelength of 1820 cm- I from a
differential spectrum relative to the solvent, the extraction procedure being superior in
terms ofboth speed and sensitivity, being able to measure FFA levels down to ~0.001%.
The method developed for moisture determination involves extracting water in edible oils
into dry acetonitrile and then quantitating it by measuring the absorbance of the OH
stretching band (3629 cm-I) and/or the HOH bending band (1631 cm- I
). AH three
methods were validated by standard addition experiments, evaluated for potential
interferences, and, in the case of FF A determination, compared to the performance of
AOCS official methods. The results indicated that the extraction-based procedures were
superior to conventional wet chemical methods in both sensitivity and reproducibility.
The FF A and H20 extraction procedures were subsequently automated by connecting an
auto-sampler to the FTIR spectrometer and developing procedures and software
algorithms to enable the analysis of up to 100 samples/h. The methods developed and
implemented are a substantive improvement over conventional methods for the analysis
of FF A and HzO in edible oils and provide a means by which QC and process
laboratories can analyze large volumes of edible oils for these two important parameters.
RESUME
Trois nouvelles méthodes d'analyse quantitative utilisant la spectroscopie à
infrarouge à transformation de Fourier (FTIR) ont été développées afin d'analyser les
huiles comestibles; deux méthodes pour mesurer les acides gras libres (AGL) et une
méthode pour mesurer l'humidité (H20), ces méthodes ont été enfin automatisées et
mises en application sur un spectromètre FTIR équipé d'un auto échantillonneur.
Les deux méthodes développées pour la détermination des AGL transforment les
AGL en leur sels de carboxylate par le moyen de la réaction acide/base sans causer de
saponification à l 'huile, la première méthode utilise le propanol, un solvant miscible avec
l'huile alors que la deuxième méthode utilise le méthanol, un solvant immiscible avec
l'huile et dans lequel les sels des AGL sont extraits. La première méthode fait appel à la
division des échantillons d'huile en deux moitiés, une moitié est traitée par une base
dissous dans le propanol et l'autre moitié est traitée avec du propanol uniquement. Le
spectre de chaque moitié est enregistré et un spectre différentiel est obtenu permettant
ainsi la quantification. La méthode utilisant le méthanol fait simplement appel à
l'extraction des AGL dans une solution de méthanol contenant une base faible et à la
quantification des sels des AGL ainsi produits. Les deux méthodes déterminent le contenu en AGL en mesurant l'absorbance du v (COO) à 1570 cm- I respectif à une
longueur d'onde de référence de 1820 cm-1 à partir d'un spectre différentiel respectif au
solvant, elles sont capables de mesurer des niveaux des AGL aussi bas que ~0.001 %. Une
méthode pour la détermination de l'humidité a éte développée en extrayant l'eau dans les
huiles comestibles dans de l'acetonitrile sec et le quantifiant par la suite en mesurant
l'absorbance de la bande d'élongation OH (3629 cm-I) et/ou de la bande de déformation
HOH (1631 cm- I). Les trois méthodes furent ensuite validées grâce à l'utilisation de
techniques d'addition standard, evaluées pour les interférences potentielles et dans les cas
de la détermination des AGL comparées à la performance des méthodes officielles de
l'AOCS. Les résultats ont montré que les procédures basées sur l'extraction étaient
supérieures aux méthodes chimiques conventionnelles en termes de sensibilité et
reproductibilité. Les méthodes d'extraction des AGL et du H20 furent ensuite
automatisées en connectant un auto-echantillonneur au spectromètre FTIR et en
développant des procédures et des algorithmes pour logiciels qui facilitent l'analyse
jusqu'à 100 echantillons/ho Les méthodes développées et mises en application constituent
une amélioration considérable par rapport aux méthodes conventionnelles d'analyse des
AGL et H20 dans les huiles comestibles et fournissent un moyen par lequel le CQ et les
laboratoires des processus peuvent analyser de larges volumes d'huiles comestibles pour ces deux importants paramètres.
11
ACKNOWLEDGMENTS
l would like to express my deepest gratitude to Professor F. R. van de Voort, my
supervisor, for the time, guidance and encouragement that he has given me throughout
my PhD program; his exceptional qualities as a research supervisor have helped make
this thesis possible. l wish to acknowledge my sincere gratitude to Dr. Jacqueline
Sedman for her collaboration, valued advice and input and to thank Drs. William
Marshall, Selim Kermasha, Varoujan Yaylayan and Ashraf Ismail for their guidance,
advice and use of analytical instrumentation.
Very special thanks to Sensei John Kalaidopoulos, my senior karate teacher,
who taught me in practice the true meaning of karate, for his support and understanding.
Very warm thanks go to Sempai Stephen, Sempai Tony, my best friend Denies and aIl
the karate students at the West Island Kyokushin Karate for their kindness, friendship
and continuous help. l also wish to extend my warm thanks to my colleagues at the Food
Science and Agricultural Chemistry Department and to my friends in Sainte Anne de
Bellevue.
Above aIl, my deepest gratitude goes to my parents, family and my beloved
woman for their continuous love, support and encouragement, with my personal thanks to
Sultan Qaboos University (SQU) and Embassy of Sultanate of Oman, Washington DC,
for the financial and academic support that they given me throughout my study in
Canada.
iii
CONTRIBUTIONS OF AUTHORS
Chapters 3-6 of this thesis comprise of the text of published or submitted papers
listed below. The author of this thesis was responsible for the concept, design of
experiments, experimental work, and manuscript preparation. Dr. van de Voort is thesis
supervisor and had direct advisory input as the work as it progressed. Dr. Sedman
provided advice relative to spectral interpretation, chemometrics and editorial assistance
in the final stages of paper submission. Mr. Ghetler was responsible for the UMPlRE
software platform on which FF A and H20 algorithms were developed for the automated
methods discussed in Chapter 6.
List of the publications reported in the thesis:
AI-Alawi, A., van de Voort, F. R and Sedman, J. New FTIR Method for the
Determination of Free Fatty Acids in Oils. J. Am. Oil Chem. Soc., 81(5), 441-446 (2004).
AI-Alawi, A., van de Voort, F. R and Sedman, J. A New FTIR Method for the Analysis
ofLow Levels ofFFA in Refined Edible Oils. Spectrosc. Lett. 38(4-5) 389-403 (2005).
AI-Alawi, A., van de Voort, F. Rand Sedman, J. A New FTIR Method for the
Determination of Low Levels of Moisture in Edible Oils. Appl. Spectrosc., 59(10), 1295-
1299 (2005).
AI-Alawi, A., van de Voort, F. R, Sedman, J. and Ghetler, A., Automated FTIR Analysis
of Free Fatty Acids and Water in Edible Oils. Journal of the Association for Laboratory
Automation. (in press).
IV
CONTRIBUTIONS TO KNOWLEDGE
1. Designed new analytical approaches for edible oil analysis to provide the
industry with practical and readily implemented methodology.
New analytical strategies were developed to compensate for matrix effects arising
in FTIR analysis of edible oils that are not readily modeled by chemometrics,
thereby enhancing analytical accuracy and sensitivity, to facilitate sample
handling, and to allow automation of the analysis.
2. Developed the first quantitative FTIR method for FF A analysis in edible oil
that avoids saponification and is independent of oH type.
3.
Quantification was successfully achieved through the use of carefully selected
weak bases which converted FF A to their salts without causing saponification of
the oil. Generalization of the FTIR method so that matrix effects were eliminated
and the spectral response was independent of oil type was achieved via
differential spectroscopy.
Developed a sensitive and simple method for determination of low levels of
FF A in refined edible oils.
The sensitivity of the FF A method was extended by treating oils with an oil
immiscible solvent containing a weak base to convert the FF A to their carboxylate
salts and concentrate the salts in a small volume. This procedure greatly enhanced
the sensitivity of the FTIR FF A analysis as well as eliminating the requirement to
prepare two samples to obtain one analytical result as in the first method devised.
4. Developed a sensitive, simple and accurate method for the quantitative
determination of moisture content in edible oils.
An FTIR-based instrumental method for the determination of moi sture was
developed, modeled on the extraction procedure developed for low-FF A
determination, but without any requirement to carry out a stoichiometric reaction.
The method eliminates most of the analytical problems that are commonly
associated with moisture quantification in edible oils and provides a viable
alternative to the widely used Karl Fischer method.
v
5. Automated the extraction-based FF A and moisture analysis methods to carry
out high-speed, high-volume analyses of these measures for QC and process
laboratory purposes.
The extraction-based FF A and moisture methods developed were refined and
implemented to operate on an FTIR integrated with an auto-sampler and micro
pump to allow for automation of these methods. The system was developed,
programmed, optimized and validated and was shown to be able to analyze up to
100 samples/h.
VI
TABLE OF CONTENTS
ABsTIti\<:T-------------------------------------------------------------------------------------------i
~SlJ~E----------------------------------------------------------------------------------------------ii
AC~O~LEI>(;~ENTS------------------------------------------------------------------------iii
<:ONTRIBlJTIONS OF A lJTH ORS ----------------------------------------------------------- iv
<:0 NTRIBlJTI ONS TO~ 0 ~LEI>(; E ------------------------------------------------------ v
TABLE OF CONTENTS------------------------------------------------------------------------ vii
LIST OF FI(;lJRES --------------------------------------------------------------------------------x
LIST OF TABLES-------------------------------------------------------------------------------- xii
LIST OF ABBREVIATI 0 NS ----------------------------------------------------------------- xiii CHAPTERI
(;ENEIti\L INTROI>lJCTION------------------------------------------------------------------- 1 1.1. CONTEXT OF THE RESEARCH ---------------------------------------------------- 1
1.2. RA TIONALE OF THE WORK -------------------------------------------------------- 3
1.3 . REFERENCES --------------------------------------------------------------------------- 4
<:HAPTER2
LITEIti\TlJ~ ~VIE~ -------------------------------------------------------------------------5 2.1. FREE F ATTY ACIDS (FF A) ANAL YSIS------------------------------------------- 5
2.1.1. Introduction ------------------------------------------------------------------------- 5
2.1.2. Historical Perspective-------------------------------------------------------------- 6
2.1.3. Colorimetric Methods-------------------------------------------------------------- 6
2.1.4. Electrochemical Methods ------------------------------------------------------- Il
2.1.5. Thermometric Methods ---------------------------------------------------------- 14
2.1.6. CUITent Status of FF A Analysis ------------------------------------------------ 14
2.1.7. Infrared Spectroscopic Methods ----------------------------------------------- 15
2.2. MOISTURE CONTENT ANAL YSIS ---------------------------------------------- 17
2.2.1. Introduction ----------------------------------------------------------------------- 17
2.2.2. Karl Fischer Method ------------------------------------------------------------- 19
2.2.3. Infrared Methods ----------------------------------------------------------------- 27
2.3. CONCLUSION ------------------------------------------------------------------------- 31
2.4. REFERENCES ------------------------------------------------------------------------- 31
vu
CHAPTER3
NEW FTIR METHOD FOR THE DETERMINATION OF FREE FATTY AC ID IN
OI~~ ------------------------------------------------------------------------------------------------- 39
3 .1. AB~TRA CT ----------------------------------------------------------------------------- 39
3.2. ~TFt()l)lJCTI()~---------------------------------------------------------------------- 39
3.3. EXPEFtIME~T AL PFt()CEl)lJFtE -------------------------------------------------- 42
3.4. FtESlJLTS A~I) I)ISClJSSI()~ ----------------------------------------------------- 44
3.4.1. Analytical Concepts-------------------------------------------------------------- 44
3.4.2. Calibration and Stability of the Fteaction ------------------------------------- 45
3.4.3. Evaluation and Validation------------------------------------------------------- 48
3.5. A CKN () WLEI)GME~T -------------------------------------------------------------- 52
3. 6. FtEFEFtE~ CES ------------------------------------------------------------------------- 52
lJRID(;E---------~----------------------------------------------------------------------------------- 54
CHAPTER4
A NEW METHOD FOR THE ANA~Y~I~ OF ~OW ~EVE~~ OF FFA IN
REFINED EDIlJ~E OI~~----------------------------------------------------------------------- 55
4.1. ABSTRACT ----------------------------------------------------------------------------- 55
4.2. ~TFt()l)lJCTI()~---------------------------------------------------------------------- 55
4.3. METH()I)() L()G Y --------------------------------------------------------------------- 59
4.4. FtESlJLTS A~I) I)ISClJSSI()~ ----------------------------------------------------- 60
4.4.1. Calibration ------------------------------------------------------------------------- 63
4.4.2. V alidation -------------------------------------------------------------------------- 64
4.5. C() ~ CL lJSI () ~ ------------------------------------------------------------------------- 69
4.6. ACK~()WLEI)GME~T -------------------------------------------------------------- 69
4.7. FtEFEFtE~CES ------------------------------------------------------------------------- 69
lJRID(;E--------------------------------------------------------------------------------------------- 72
CHAPTER5
A NEW FTIR METHOD FOR THE DETERMINATION OF MOI~TURE IN
EDIlJ~E OI~~------------------------------------------------------------------------------------- 73
5.1. ABSTRACT ----------------------------------------------------------------------------- 73
5.2. ~TFt()l)lJCTI()~---------------------------------------------------------------------- 73
Vlll
5.3. 1V1~lllI()I)()~()CJ): --------------------------------------------------------------------- 75
5.4. RESU~llS ANI) I)ISCUSSION ----------------------------------------------------- 78
5.4.1. CJeneral Considerations ---------------------------------------------------------- 78
5.4.2. Calibration I)evelopment -------------------------------------------------------- 80
5.4.3. Consideration of Interfering Constituents ------------------------------------ 82
5.4.4. Factors Affecting Analytical Accuracy --------------------------------------- 84
5.5. A CKN () W~~I)CJ~1V1~NllS----------------------------------------------------------- 85
5 .6. REF~RENC~S ------------------------------------------------------------------------- 85
1JFlII>(;IG--------------------------------------------------------------------------------------------- 87 CHAPTIGR6
AUTOMATIGI> FTIR ANALYSIS OF FRIGIG FATTY ACII>S OR MOISTURIG IN
IGI>I1JLIG ()ILS------------------------------------------------------------------------------------- 88
6.1. ABSllRACll ----------------------------------------------------------------------------- 88
6.2. INllFt()I)UCllI()N---------------------------------------------------------------------- 88
6.3. 1V1~lllI () I)() ~OCJ): --------------------------------------------------------------------- 88
6.4. RESU~ Ils ANI) I)ISCUSSI()N ----------------------------------------------------- 92
6.4.1. System ()verview and Software ------------------------------------------------ 93
6.4.2. System ~valuation and Validation --------------------------------------------- 99
6.5. C()NC~USI()N ------------------------------------------------------------------------ 101
6.6. A CKN () W~~I)CJ 1V1~Nll ------------------------------------------------------------- 1 02
6.7. REF~~NC~S ------------------------------------------------------------------------ 102
CHAPTIGR 7
(;IGNIGRAL CON CL USI ON ------------------------------------------------------------------- 104
IX
LIST OF FIGURES
Figure 2.1. Schematic diagram of the copper soap method for FF A analysis introduced
by Baker in 1964 ...................•...............•.•........•..........•............•...•..........•...•.... 7
Figure 2.2. Flow diagram of the FIA instrument for determination of FF A in oi! samples
as proposed by Ekstrom and Rhee .....................................................•............. 8
Figure 2.3. Flow diagram of the FIA instrument for determination of FF A in oil samples
as proposed by Canham .•.....•.•...••.•.•.•.••.•.•.•.••••.•..•.•..••.••••••.••••••..••••.••..••••••.•...• 9
Figure 2.4. Flow diagram of the FIA instrument for determination of FF A in oil samples
as proposed by Linarers et al. in 1989 ........................................................... 11
Figure 2.5. Metrohm auto-titrator instrument equipped with autosampler for FF A
analysis .......................................................................................................... 13
Figure 2.6. Shift in FF A band after conversion to carboxylate salt. FF A shows band at
1711 cm-1 (A); the band shifts to 1570 cm-1 after addition of base (B) ..•...... 16
Figure 2.7. FTIR spectrum ofpure water .......................•..•.......•..•.••.••.•.••.•..•.•••.•.•.•.•.••.•.• 28
Figure 3.1. Schematic diagram illustrating the sample preparation procedure and the
steps in the analytical protocolleading to the differential spectrum ...•...•..•.• 43
Figure 3.2. Potassium phthalimide as a reagent to carry out the acid/base reaction ........ 46
Figure 3.3. Composite calibration curve obtained from the differential spectra of three
independent sets of hexanoic acid-spiked soybean oi! standards of the type
illustrated in Figure 3.1 ..............................................................•.•.....•...•..•.•. 47
Figure 3.4. Graphical comparison of results obtained by the AOCS and FTIR methods
for the oils that were subjected to standard addition of FF A mixture, relative
to the gravimetrical data ................................................................................. 50
Figure 4.1. Schematic diagram of the sample preparation and analytical protocol. ........ 61
Figure 4.2. DifferentiaI spectra for sodium hydrogen cyanamide in methanol.. .............. 62
Figure 4.3. DifferentiaI spectra for soybean oil spiked with palmitic acid (0.0 - 0.1 %)
after carrying out the acidlbase reaction ....•..•.........................•..•................... 64
Figure 4.4. Calibration curve for %FF A in oïl obtained from the differential spectra in
Figure 4.2. The %FFA is expressed as % oleic acid (w/w) ........................... 65
Figure 5.1. Schematic diagram of the FTIR sample handling system ............................. 76
x
,~,
Figure 5.2. Schematic diagram of the sample preparation procedure for moisture analysis
of edible oils and the FTIR spectral analytical protocol. ............................... 77
Figure 5.3. Spectra of water/acetonitrile standards (0, 400, and 800 ppm added water)
(upper series) and the corresponding spectra obtained after subtraction of the
spectrum of the acetonitrile employed to prepare the standards (lower series) .
....................................................................................................................... 79
Figure 5.4. DifferentiaI spectra obtained after extraction of water from water/canola oil
standards (0-2000 ppm water) into acetonitrile, recording the spectrum of the
acetonitrile layer, and ratioing the spectrum against that of the acetonitrile
extraction sol vent ........................................................................................... 81
Figure 5.5. DifferentiaI spectra of acetonitrile spiked with hexanol (A), glycerol (B),
tert-butyl hydroperoxide (C), oleic acid (D), and water (shaded) .............•..• 83
Figure 6.1. General analytical protocol for FF A and H20 analysis of edible oils by FTIR
spectroscopy .................................................................................................. 92
Figure 6.2. The FTIRIauto-sampler system used for the analysis ofFFA or moisture .... 94
Figure 6.3. Auto-sampler needle punching through the Mylar liner. ............................... 94
Figure 6.4. The user interface of UMPIRE® Pro pro gram. A and B, the "General" and
"Scanning/Sampling" tabs for the "Modify Methods" submenu, respectively;
C, "Peak/Area Position" tab for the "Components" submenu ....................... 95
Figure 6.5. DifferentiaI spectra of canola oil standards spiked with water (0-2000 ppm)
after extraction with acetonitrile .................................................................... 98
Figure 6.6. DifferentiaI spectra of canola oil standards spiked with palmitic acid (0.0-
0.1 % w/w) after extraction with methanol containing NaHNCN .•..•........•.•.. 98
Figure 6.7. Plots for moisture content vs. amounts added for the validation run ........... 100
Figure 6.8. Plots for FF A vs. amounts added for the validation run .............................. 101
Xl
LIST OF TABLES
Table 1.1. Data Entry Report for Cottonseed Oil (2001-2002) ........................................ 2
Table 3.1. Results of Triplicate Analyses of Oils Spiked with Various Amounts an FF A
Mixture by the AOCS Titrimetric and FTIR Methods .................................. 49
Table 3.2. Results of Triplicate Analysis of Smalley Check Samples by AOCS Method
and FTIR Method .......................................................................................... 51
Table 4.1. FTIR Analyses Based on the Use of Stoichiometric Reactions for "Signal
Transduction" ................................................................................................ 57 .
Table 4.2. Mean Difference (MD) and Standard Deviation of the Difference (SDD) for
Accuracy (a) for the Titrimetric and FTIR Procedures vs. Gravimetrie
Addition and for Reproducibility (r) for Duplicate Analyses Carried Out by
Titrimetric and FTIR Analyses, Respectively ............................................... 67
Table 4.3. Results of Triplicate Analyses of Locally Purchased Oil Samples by the
IUPAC Reference (Titrimetric) Method and the FTIR Method .................... 68
Table 6.1. FF A and Moisture Content of Various Edible Oils (mean of triplicate
analyses). . ...................................................................................................... 99
Xll
LIST OF ABBREVIATIONS
I-PrOH n-Propanol
AN Acid number
AOAC Association of Official Analytical Chemists
AOCS American Oil Chemists' Society
ASTM American Society for Testing and Materials
ATR Attenuated total reflectance
BR Blank reagent
DMP Dimethoxypropane
DMSO Dimethyl sulfoxide
DTGS Deuterated triglycine sulfate
EDTA Ethylenediaminetetraacetic acid
FFA Free fatty acid
FIA Flow injection system
FTIR Fourier transform infrared spectroscopy ~--
IUPAC International Union of Pure and Applied Chemistry
K-Phthal Potassium phthalmide
LIMS Laboratory information management system
MD Mean difference
MeOH Methanol
MLR Multiple linear regression
NIR N ear infrared
NSERC Natural Sciences and Engineering Research Council
OD Optical density
PLS Partialleast-squares regression
ppm Parts per million
PV Peroxide value
QC Quality control
RBD Refined, bleached and deodorized oil
/'~--" ROOH Hydroperoxide
RR Reactive reagent
Xll1
,/ -, ~ SD
SDD
SEFIA
TAG
TEk
TPP
TPPO
UMPIRE®
w/v
w/w
ZnSe
Z-reg
Standard deviation
Standard deviation of the differences
Solvent extraction flow injection system
Triacylglycerol
Triethanolamine
Triphenylphosphine
Triphenylphosphine oxide
Universal Method Platform for InfraRed Evaluation
weight per volume
weight per weight
Zinc selenide
Regression through the origin
XIV
r·· CHAPTERI
GENERAL INTRODUCTION
1.1. CONTEXT OF THE RESEARCH
Fats and oils represent a major and essential portion of our diet, and the fats and
oils industry is one ofthe large st sectors of the food industry. According to Euromonitor
International, consumer expenditure for fats and oils in 2005 was US $ 122.6 billion (1),
comprising - 0.5% of the total consumer expenditures; this amount does not include by
products such as soap and meal cake. Because fats and oils are valuable and economically
important food commodities, a variety of professional organizations have set standards
for the edible oil industry in relation to determining their quality, functionality and the
parameters affecting their economic value. Several prominent organizations vie to play
official roles in this regard, including the American Oil Chemists' Society (AOCS) (2),
the Association of Official Analytical Chemists (AOAC) and the International Union of
Pure and Applied Chemistry (IUP AC) (3). Methods and standards adopted by these
organizations are considered to be "official" methods, most being similar, albeit not
identical in their detail. A wide range of official analytical methods exist for the routine
evaluation and characterization of fats and oils, examples being iodine value (degree of
unsaturation), saponification value (average molecular weight), moisture content,
peroxide value, anisidine value and free fatty acid (FF A) content (4), to name just a few.
Most of the official procedures noted above are titrimetric methods developed in
the 1930-40's using color indicators and many are still in routine use today. AlI are
workable and are standardized; however, the lack of skilled labor and the
safety/environmental issues associated with toxic solvents and reagents make these
methods problematic in modem times. Apart from these drawbacks, most of these
methods are mediocre in terms of accuracy, reproducibility and analytical range relative
to newer instrumental methods. The po or accuracy is often associated with their non
specificity, the reliance on accurate color perception by the analyst or problematic dark
samples, the latter obscuring the end point. Table 1.1 shows an example of an inter
laboratory study of FF A and moi sture results obtained for the same sample of cottonseed
oil (5) as reported by 10 different certified laboratories using the same official methods.
1
Table 1.1. Data Entry Report for Cotionseed Oil (2001-2002). Adopted from ref. (5)
Analyst FFAa Moistureb
1 0.14 0.10
2 0.15 0.05
3 0.16 0.09
4 0.15 0.10
5 0.07
6 0.04 0.09
7 0.17 0.04
8 0.05 0.08
9 0.18 0.06
a % w/w expressed as oleic acid; b % w/w
Such variations can increase even further as a result of biases which may
interpose themselves when "official" methods from different organizations are used (e.g.
AOCS vs. IUPAC), as each has different procedure. For environmental, sample
throughput and accuracy reasons, there has been an ongoing effort to develop new
instrumental approaches to titrimetric based analyses, focusing on automated titrators and
continuous flow injection systems using various potentiometric and colorimetric
detection systems.
An alternative to titrimetric analyses is the application of spectroscopic techniques
to determine oil quality parameters based on spectral characteristics of the sample. In
particular, a wide range of oil parameters can now be analyzed by FTIR spectroscopy (6),
providing alternative approaches to their assessment. The Mc Gill IR Group has been at
the forefront of the development of instrumental methods for the analysis of edible oils
based on Fourier transform infrared (FTIR) spectroscopy. Whereas infrared spectroscopy
.~... was generally regarded in the past as a qualitative technique, providing functional group
information, it has now evolved into a quantitative spectroscopie procedure, supported by
2
powerful chemometrics. Thus quantitative work with FTIR instrumentation can now
approach the simplicity of using UV/visible spectroscopy, while providing much more
specific spectral information in relation to the molecular species targeted for analysis.
However, for complex samples such as edible oils there are certain inherent limitations to
FTIR analysis such as matrix effects, overlapping bands, hydrogen bonding effects, etc.
Chemometric techniques such as partial least squares regression can overcome sorne of
these issues, but not aIl of them. Method development by the McGill IR Group has
undergone an analytical evolution which exploits a range of strategies to obtain ever
better sensitivity, accuracy or speed. The ultimate goals are to improve sensitivity and
accuracy and ultimately to standardize and automate edible oil analysis by FTIR
spectroscopy.
1.2. RATIONALE OF THE WORK
Of the variety of methods one could tackle with the above goals in mind, FF A
analysis was my personal choice, based on the thousands of titrations which 1 had the
misfortune to carry out in an oil processing plant, being a good motivator. Although FTIR
methods for FF A determination have been previously developed, they suffer from several
shortcomings. Similarly, there are weIl known problems associated with the titrimetric
determination of moi sture by the Karl Fischer method and this provided the impetus for
developing methodology for the determination of moisture in edible oils by FTIR
spectroscopy. Both FF A and moisture analyses are used to control many steps in the
overall oil refining process, such as deodorizer efficiency and soap adsorption, and to
evaluate storage conditions and oil stability. Thus these analyses determine important
quality and process control parameters which need to be determined accurately and
precisely. An examination of the literature regarding the current status of FFA and
moi sture analysis, indicates that there is still substantial room for improvement in terms
of accuracy, specificity and practicality as weIl as a need for methods that can be
automated by QC and process laboratories. Chapter 2 of this thesis reviews and details
the relevant analytical methodology and approaches to the analysis of FF A and moisture
to pro vide a context for the FTIR methods developed. In Chapter 3, a' method for the
/~. determination of FF A based on a stoichiometric reaction coupled with differential
spectroscopy is presented. Chapter 4 revisits the method, considering extraction as a
3
,~
means of increasing sensitivity. Chapter 5 describes a method developed for moisture
analysis drawing in part on the successful approach used in Chapter 4 for FF A. Chapter 6
takes latter two methods and adapts and implements them on an auto-sampler equipped
FTIR to automate the methods attaining analysis rates of ~ 100 samples per hour as weIl
as testing them to determine their overall performance. Each chapter is presented as a
paper, Chapters 3-5 having been published, with Chapter 6 in preparation for submission.
It has been a challenge and learning experience to develop methodology which 1
personally feel will eventually impact the edible oil industry in a tangible way in the
future.
1.3. REFERENCES
1.
2.
3.
4.
5.
6.
Consumer expenditure on oils and fats: National statistical offices/OECD/EurostatlEuromonitor International
AOCS, Official Methods and Recornrnended Practices of the Arnerican Oil Chernists' Society, 4th edn., AOCS Press, Champaign, IL, 1998.
IUPAC, Standard Methods for the Analysis of Oils, Fats and Derivatives, 7th edn., Blackwell Scientific, London, 1992.
O'Brien, R. D., Fats and Oils: Forrnulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998.
Laboratory Proficiency Pro gram, Cottonseed oil, AOCS, Champaign, IL, 2002.
van de Voort, F. R., Sedman, J. and Russin, T. Lipid Analysis by Vibrational Spectroscopy, Eur. J. Lipid Sci. Technol. 103,815-826 (2001).
4
CHAPTER2
LITERATURE REVIEW
2.1. FREE FATTY ACIDS (FFA) ANALYSIS
2.1.1. Introduction
Free fatty acids (FF A) are common triglyceride hydrolysis products found in
crude oils and to sorne extent in refined oils as weIl as developing as a result of oxidation
or frying, ultimately impairing oil quality and functionality. ChemicaIly, FFAs are less
stable than triglycerides and therefore more likely to oxidize and cause ranci dit y (1).
From a quality standpoint, the FF A content of edible oils is not only a crucial factor
associated with their quality and economic value, but also an important quality indicator
in relation to fats and oils processing, i.e., serving as a measure of deodorizer efficiency
(1). The AOAC, AOCS, and IDPAC aIl have standard methods for the determination of
FF A, with the AOAC and AOCS methods being identical. These methods are based on
the titration of an oïl (3.5-56.4 g) dissolved in hot neutralized 95% ethanol (50-100 ml)
(2) or ethanol/diethyl ether (50-150 ml) (3) with a strong base to a phenolphthalein end
point under vigorous shaking. Although simple, these titrimetric methods are tedious and
consume substantial amounts of solvent which are costly and difficult to dispose of
safely; in addition, the use of a colorimetric indicator is problematic when dark crude oils
are analyzed.
In recent years, a variety of approaches have been investigated as possible
alternatives to the manual titrimetric methods employed to determine the FF A content of
oils. These generally require more sophisticated equipment such as flow injection
systems, automatic titrators and chromatographic instrumentation, among others. Many
of these techniques offer sorne benefits in terms of speed of analysis, amenability to
automation, and/or a reduction in the use of solvents, with the attendant environmental
benefits, and sorne provide a substantive gain in sensitivity over that attained with
titrimetric methods. The objective ofthis section is to review in general terms a variety of
approaches to FF A analysis and place the development of FTIR methodology in context.
5
2.1.2. Historieal Perspective
After W orld War II, the food industry began a new stage of development and
expansion to catch up with the increased need for food as a result of the growing
population. As an integral part in the food chain, the oil industry expanded dramatically
which led to revisionlre-evaluation of many quality associated parameters and the ,
methods that are used to determine them. FF A determination was among the important
methods that were subjected to many studies and modifications to make it more sensitive
and reproducible as weIl as more rapid and economical. Over the past sixt Y years, many
methods were published, and each method utilized the analytical technology advances
associated with its time to improve the method as weIl as addressing the health and
environmental issues of concern at the time of publication. For example, as
spectrophotometry was the most weIl known analytical technique in the 1950s, most of
the FFA methods published at that time (and later) were colorimetric methods as opposed
to the previous generation of methods which were characterized by the use of manual
titration for quantification. Later on, the methods tended more towards electrochemical
techniques.
2.1.3. Colorimetrie Methods
Copper soap method. The first attempt to develop a colorimetric method for FF A
was carried out by Ayers in 1956 (4). FF As were converted into water-insoluble soaps by
reacting them with an aqueous solution of copper nitrate or cobalt nitrate. Quantification
was carried out by extracting the colored soap precipitate (blue for copper and pink for
cobalt) in chloroform followed by measurement of absorbencies at 675 nm for copper and
527 nm for cobalt. Although the method was not intended for FF A analysis per se, it
provided a basis for other researchers to develop colorimetric methods for FF A analysis
in edible oils.
In 1964, Baker (5) introduced his improved version of the cupric method which
was designed especially for FF A analysis in edible oils. The method involves dissolving
the oil sample (8 g) in benzene (50 mL) followed by a mixing step where an aliquot (10
mL) of the mixture is shaken with aqueous cupric acetate solution in a separate tube. The
mixture is centrifuged or let settle until a clear separation between the two phases is
6
obtained. Quantification of FF A is determined by measuring absorption of the upper
benzene layer at 640 nm. A schematic diagram of the procedure is shown in Figure 2.1.
Although Baker claimed a good correlation between his method and the titrimetric
method, he indicated that certain copper soaps such as copper salts of palmitic and stearic
acids do not dissolve in benzene. Discrepancies between the parameters used by Barker
and Ayers, such as the working wavelength (640 vs. 675 nm), led Bains et al. (6) in the
same year to examine the spectra and the effect of oil color on the measurement. The
author concluded that 670 nm is the optimum wavelength, giving high sensitivity for
quantification of copper soaps in colored or non-colored oil.
jiP 11 ----11> .. ;:
Benzene + Où
,",
r---+-----, Aqueous Cupric Acetate '---_S_hak_e -', solution
~ Centlifuge 1
~
i JI Spectrophotometer
À= 640nm
Figure 2.1. Schematic diagram of the copper soap method for FF A analysis introduced by Baker in 1964 (5).
Despite the simplicity of the method, not much work was done to understand the
chemistry of the products formed or even to optimize the conditions of the method to
make it more sensitive and reproducible. In 1976, Lowry and Tinsley (7) published a
considerable work about the conditions and chemistry involved in order to provide a basis
to optimize the method and make it more sensitive and reproducible. The outcome of
7
their work was a stabilization of the copper soaps of saturated FF A in benzene as weIl as
an increase in the sensitivity of the measurement by a factor of 10 just by controlling pH
by adding pyridine. Due to both health and practical concems, a number of publications
came out suggesting a replacement for benzene, including toluene (8), isooctane (9) and
·cyc1ohexane (10).
In 1981, Ekstrom and Rhee (8) introduced the first automated system for FF A
analyses in edible oils using the copper soap method which was able to analyze up to 20
samples/h. Automation was achieved using a flow injection analysis (FIA) system to mix
oil samples with the solvents and reagents used in the method. According to the setup of
the system, separation of the organic phase from the aqueous one was achieved online
through the use of segmentors and separators at specified locations in the system. In
contrast with the previous versions of the method, the measurement was done in the
aqueous phase instead of the organic phase. This was possible by implementing two
mixing-separation steps in the system; the first one was responsible for mixing the oil
solvent mixture and cupric solution together to produce FF A-copper soaps and then
separate the two phases from each other (Figure 2.2). Wasle
Waste
-e- Pump l:x:xl MixingCoit
I:§J Injection Valve ~ Separator
• Segmentor
~~ Figure 2.2. Flow diagram of the FIA instrument for determination of FF A in oil samples as proposed by Ekstrom and Rhee (8).
8
The second step involved extracting the copper ions in FF A-copper soaps into
another aqueous solution containing EDTA; and then the copper-EDT A complex formed
was measured spectrophotometrically at 660 nm. The system did not receive much
attention, mainly because it was complicated and had stability problems. Six years later,
Canham and Pacey (11) reported an improved FIA system using a one-extraction-step
process and improved hardware (segmenter and separator) illustrated in Figure 2.3. The
new system had improved stability and shorter analysis time (130 samples/h), but at the
expense of sensitivity. The sensitivity of FIA systems was boosted by Puchades et al. in
1994 (12) mainly by optimizing operating conditions, such as flow rate, pH, tubing
length/diameter, and oil/solvent proportion, while maintaining good analysis speed (up to
75 samples/h)' Two years later, Zhi et al. (13) came up with a new solvent extraction FIA
(SEFIA) setup and protocol for FF A analysis without requiring a segmenter or separator.
The method had the same high sensitivity as the previous method, but the analysis time
was significantly increased (12 samples/h) and the FIA system required substantial
coordination and timing which was provided by a minicomputer.
Waste
-®- Pump 1 :::>c:>c 1 Mixing Co.il
t:§:l Injection Valve Restriction Coi1
• Segmentor Separator
Figure 2.3. Flow diagram of the FIA instrument for determination of FF A in oil samples as proposed by Canham (11).
9
Despite the advantages offered by the copper soap method (limit of detection
~0.01 %FFA) and its automated versions (~12-130 samples/h), the method did not
receive much attention from the oïl industry, mainly because of the complexity of the
analysis compared to the official method. Moreover, the FIA systems employed in the
analysis had specialized hardware (segmenters and separators) which made them not
suitable for other analyses.
Phenolphthalein based method Despite the drawbacks of the "official" method,
its simplicity and wide recognition made its automation a necessity in order to shorten the
analysis time and reduce the ambiguity in determining the end point. The first substantive
attempt to automate the official method was carried out by Linarers et al. in 1989 (14).
The setup of the system was simple (Figure 2.4) and used the same reagents as the
IUP AC official method (3) whilst quantification was based on monitoring the decrease in
the phenolphthalein intensity (at 562 nm) as a result of the reaction between KOH and
FF A in the oil sample. The system had an analytical range of 0.15-0.81 % FF A and
sample throughput of about 60 samples/h. In 1997, Nouros et al. (15) reported a
modification with a similar analytical range, the mixing coil being replaced with a mixing
chamber plus using a less expensive and less problematic solvent (l-propanol) as a
diluent and carrier. The sampling rate was reported as 30-100 samples/h, and solvent
consumption as 3-7 ml per sample. A similar system equipped with a photo probe
connected to a spectrophotometer was reported in 2001 by Mariotti and Mascini (16) for
FF A measurement in virgin olive oil. The system had the same sensitivity as the previous
ones with a sample throughput of 12-60 samples/h.
Phenol red with reverse micelles. In 1990, Walde (17,18) developed a new
approach for FF A determination in non-aqueous medium without titration. The method
relies on the acidity of FF A to cause color changes to the alkaline form of phenol red.
The indicator, which is water soluble, is dispersed in the non-aqueous medium
(oil/isooctane) by bis(2-ethylhexyl) sodium sulfosuccinate, a surfactant. Quantification
was achieved through monitoring changes in absorbance at 560 nm (disappearance of the
red color). Although this method was economical in oil and solvent usage (0.1 ml and 5
ml, respectively), it was found to be lacking in accuracy and no further attempts have
been made to optimize or automate it.
10
Waste
-®- Pump 1 )OC 1 Mixing Coil
~ Injection Valve
Figure 2.4. Flow diagram of the FIA instrument for determination of FF A in oïl samples as proposed by Linarers et al. in 1989 (14).
2.1.4. Electrochemical Methods
Voltammetric techniques. Voltammetric techniques are widely used for
compounds that are readily reduced or oxidized to provide highly selective and sensitive
quantitative methods. Although FF A can be oxidized or reduced, they are by nature
weakly active electrochemically, a factor which impeded the development of reliable
electrochemical methods for FF A determination. The situation was changed in 1972
when Takamura et al. (19) published their work on the voltammetric behavior of quinone
in the presence of acids in unbuffered protic solvents. It was reported that addition of
acids to the reduced form of quinone (dissolved in ethanol) gave a very strong
voltammetric signal. The behavior of quinone has been utilized by many authors (20-25)
to develop FIA and HPLC methods for FF A analysis in edible oils. The FIA system
coupled with voltammetric detection had high sensitivity/reproducibility (0.05 % FF A,
RSD 0.7% (23», low solvent consumption and outstanding throughput (60 samples/h)
~', pH-Metric methods. The pH meter obviously cornes to mind as a means to
measure acidity (FF A); however, the high viscosity and low polarity of oils have made it
11
impractical to use pH-metric methods. In 1991, Labshina et al. (26) suggested extraction
of FF A from oïl samples into a polar solvent to carry out pH measurements. In order to
facilitate extraction and pH measurement, the authors suggested use of a weak base,
triethanolamine (TEA), in a suitable solvent mixture (water (1 %), diethyl ether (80%) and
chloroform (19%» to convert carboxylic acids into their salts. Although workable, the
method had poor reproducibility, mainly because of instability of the electrode and the
basic reagent in the solvent mixture (due to the low concentration ofwater). The method
was later modified by Tur'yan et al. (27), who replaced the former solvent mixture with a
1: 1 (v:v) mixture of iso-propanol and water containing 0.2 M TEA and 0.02 M KN03, the
latter added to obtain a constant ionic strength. Rather than extracting FF As, this solvent
mixture effectively makes an emulsion within which the pH measurement is made.
Various versions of this approach were developed for acidity measurements in petroleum
oils (28), oilseeds (29) and hydraulic oils (30). Although providing a simple means of
measuring FF As, it does not provide any improvement over the official titrimetric method
in terms of sensitivity or solvent consumption.
Automated potentiometric titration. One of the main reasons that the official
method is still in use is the advances that have been made in potentiometric
measurements and automated titration systems. The first application of potentiometric
measurements to determine the end point of acidlbase reactions goes back to the early
years of the last century. It was realized that a potentiometric end point is sharper and
more reproducible in determining the end point of a reaction than using a color indicator;
however, a lack of the technology required to automate the measurement limited the
acceptance of this approach as a practical alternative to the conventional titration. Since
W orld War II, substantial effort has been devoted to improving potentiometric response
as well as automation and has resulted in the commercial availability of potentiometric
auto-titration systems (31-35). The basic setup of such a system consists of a
potentiometric electrode (pH meter), syringe pump, and stirrer and a recorder (or
computer). Initial automated titrators were large and had few features, but
computerization has overcome many of their initial limitations, resulting in general
acceptance in analytical labs. Figure 2.5 presents a modem potentiometric auto-titration
system which in brief, involves volumetric addition of the titrant (alkaline solution) from
12
the bottle at the rear (see Figure 2.5) in small increments by the syringe pump while the
pH is measured by the electrodes immersed in the beaker of analyte on the stirrer. By
convention, the titration curve has the volume of titrant added on the x-axis and the
corresponding change in pH (expressed as pH units or m V) on the y-axis. The end point
of the titration is determined as the midpoint of the maximum slope of the titration curve.
Once the end point is detected, the volume of the titrant added stoichiometrically defines
the concentration of FF A in the oil sample.
Figure 2.5. Metrohm auto-titrator instrument equipped with autosampler for FF A analysis.
Auto-titrators are available comlJlercially from a number of suppliers, such as
Metrohm, Mettler, Labcor, etc., and can be equipped with many optional accessories, e.g.
autosampler, different sens ors, etc. For FF A analysis, most of the titrators use glass
electrodes to measure changes in pH, al: 1 mixture of ether and etllanol as diluent, and
13
KOH/ethanol (or KOHltert-butanol) (36) as titrant. The accuracy of the auto-titrator
systems depends mainly on the capability of delivering small and consistent amounts of
the titrant, which varies between different manufacturers. The sensitivity, on the other
hand, depends more on the pH electrode; if the electrode receives proper conditioning
treatments, good sensitivity (on the order of 0.01% FFA) will be maintained.
Potentiometric auto-titrator apparatuses are the most widely used systems for FF A
analysis in quality controllabs.
2.1.5. Thermometrie Methods
Thermometric methods are titrimetric methods that use the change of temperature
resulting from a chemical reaction to determine the end point of the titration. The
principle of this type of method is the same as that for potentiometric auto-titration; the
only difference is the type of sensor (detector). The detectors (usually thermistors) used
in this kind of method show large changes in their electrical resistance for small changes
in temperature. The first considerable advance in determining the end point of the
acid/base reaction in non-aqueous medium thermometrically dates back to the 1960s. In
1965, Vaugham (37) reported that upon titration of a sample containing a weak acid
dissolved in acetone, the first excess of base catalyzes a strong exothermic reaction which
can be detected easily. Since the introduction of this method, many modifications have
been made which involve the use of different reagents and solvents in order to make the
method more sensitive and the end point sharper. In 2003 Thomas (38) reported that
acetone and chloroform react exothermically (and violently under certain conditions) in
presence of base, and performance experiments using a commercial automated
thermometric titration unit with KOH in iso-propanol as titrant showed exceptional
reproducibility in determination of FF A in edible oils. Titrations using thermistor
detectors are becoming popular in quality control labs, because these sens ors are largely
maintenance-free.
2.1.6. Current Status ofFFA Analysis
Many of the instrumental methods discussed above offer substantial benefits over
traditional manual titration in terms of speed of analysis, amenability to automation,
and/or a reduction in the use of solvents or amelioration of disposaI costs. However, the
14
colorimetric methods have inadequate sensitivity, with the additional drawback of
requiring frequent calibration, which makes them generally impractical for routine
analysis. Electrochemical methods require frequent titrant standardization, and the high
accuracy and reproducibility that these methods can provide are also dependent on the
conditions of the electrodes, making regular maintenance necessary. Moreover, none of
the methods discussed pro vide specificity to FF A measurement as they tend to be based
on measures of total acidity rather than FF A per se. Thus, there remains a need for a
specific, accurate and simple method for the analysis of FF A in edible oils.
2.1. 7. Infrared Spectroscopie Methods
The possibility of employing FTIR spectroscopy in FF A analysis has been
investigated by a number of researchers over the past fifteen years. The first such study,
reported by Lanser et al. (39) in 1991, was directed toward the development of a semi
quantitative method for the analysis of soybean oils containing appreciable levels of FF A.
The FTIR spectra of oils placed between two KBr windows were collected, and the
spectral changes observed in the carbonyl region, after deconvolution of the spectra over
the 2000-1600 cm- I range, were correlated to the FFA content obtained by the standard
AOCS method. The method was placed on a quantitative basis by devising a calibration
using standards prepared by spiking oleic acid into FFA-free soybean oil. It was noted
that because the carboxylic acid c=o absorptions appear as a broad shoulder on the
strong triglyceride ester linkage absorption, a major limitation to making the method
more quantitative was the spectral variability of the ester band among different oils. In
1993, Ismail et al. (40) developed a more rigorous quantitative method based on
measuring the carboxylic acid c=o band at 1711 cm- I after applying oil samples in their
neat form onto an attenuated total reflectance (ATR) cell. The matrix effects noted by
Lanser et al. (39) were reduced by ratioing the calibration and sample spectra against the
spectrum of an FF A-free oil of the same type as the oil being analyzed, rather than an air
background spectrum. In recognition of the limited utility of this approach in eliminating
matrix effects in the case of crude and oxidized oils (40), an alternative indirect method
was also developed. This method involved adding KOHlMeOH to the oil so as to convert
any FFAs present to their salts, effectively moving the measurement band from 1711 cm- I
to 1570 cm-l, a region where there are no major interferences which limit the accuracy of
15
peak height measurements (see Figure 2.6). In addition to overcoming matrix effects, this
indirect method was more sensitive than the direct measurement of FF A absorption,
albeit at the expense of complicating the analysis. However, the method was susceptible
to error due to saponification of the oil by the KOHlMeOH reagent.
0.3
0.2 ~ A B t:J
== ~ ,.Q .... 0 0.1 r7J
,.Q
<
0.0
-0.14-----------~------------~----------~ 1800 1700 1600 1500
Wavenumber (cm"1)
Figure 2.6. Shift in FF A band after conversion to carboxylate salt. FF A shows band at 1711 cm- l (A); the band shifts to 1570 cm- l after addition of base (B). (The spectra were taken in our lab)
After these initial studies, two publications dealt with FF A analysis by FTIR
spectroscopy in relation to palm and olive oil, respectively (41,42), both using partial
least squares (PLS) to develop relationships between spectral changes and the results
obtained with the standard method. Subsequently, Verleyen et al. (43) followed up more
rigorously on the initial work of Lanser et al using peak height measurements at 1711 cm-
1 to develop workable calibrations for a variety of oils but ultimately concluded that the
calibrations were heavily oil dependent.
The most sophisticated work related to FF A analysis has been that of Caiiada et
al. (44), who developed an elegant automated FTIR-based continuous-flow analysis
16
system capable of analyzing -40 samples/h. For this system, these authors adapted the
indirect KOH/MeOH method developed by Ismail et al. (40) as an effective means of
overcoming the matrix effects associated with direct measurement of the carboxylic acid
C=O band. Having noted the susceptibility of oils to saponification by KOH even within
the 30-s mixing time employed, Caiiada et al. modified the system to eliminate this
source of error by reducing the contact time between KOH and the oil (premixed with
MeOH) to -2 s.
Aside from these FTIR studies, FF A analysis by other types of vibrational
spectroscopy, namely, near-infrared (NIR) and Raman spectroscopy, has also been
investigated (45-49). AlI the NIR and Raman methods that have been published to date
are direct methods in the sense that there is no sample preparation involved and the FF A
measurement is done using the spectrum of the neat oil. In addition, all these methods use
PLS and/or multiple linear regression (MLR) calibrations that have been developed for
specific oil types and hence cannot be applied to other oil types or samples of unknown
origin. Moreover, the NIR and Raman methods, although having the benefit or being
reagent-free, tend to have relatively low sensitivity (~ 0.3%), which makes them suitable
largely for the analysis of oils known to contain high amounts of FF A such as fish
(45,46), palm (47), and olive oils (48, 49).
2.2. MOISTURE CONTENT ANAL YSIS f '1.
, . 2.2.1. Introduction
Moisture content in fats and oils is considered a key quality parameter in the
edible oil industry and has to be controlled throughout oil processing, this being
especially important when oils are being pressed from high-moisture raw materials.
Moreover, water is commonly added to oils during degumming and refining operations
which then needs to be removed, usually by centrifugation and/or vacuum drying.
Attaining a low moi sture content after refining is very important for the effective
adsorption and removal of traces of soap in the oil (2).
17
Water has limited solubility in oils and fats, ranging from 0.05 to 0.3% (50);
however, from a quality standpoint, the moi sture content should be reduced to the lowest
practical level. By convention, moisture content of dried refined oil should be below
0.1% and is most often on the order of 0.05% (1). Hydrolysis, breakdown of fat, is
induced by the presence of moi sture, accelerated by heat and/or residual enzymes, and
results in the formation of FF A, di- and monoglycerides and glycerol. HYdrolysis can
result in off-flavors, a reduced smoke point, increased fat absorption and darkening of
shortenings used for frying (1). Maintaining a low moisture content williimit hydrolysis
of triglycerides during processing and storage (1) and avoid the formation of FF A, which
are susceptible to auto-oxidation. Thus, the routine determination of moisture would be
helpful in ensuring acceptable oil quality; however, moi sture analysis in oils is both
difficult and problematic.
There are many classic methods for the determination of moisture in edible oils
and related products, but they can be grouped into three categories: evaporation,
distillation and titration procedures. The Official Methods of the American Oil Chemists'
Society (51) lists several evaporation procedures which involve heating oil samples on a
hot plate (Ca 2b-38), in an oyen (Ca 2c-25), or in a vacuum oyen (Ca 2d-25). Although
these procedures are simple, there are many problems associated with them. For example,
the evaporation of moi sture may also remove other volatile components, such as short
chain FF As, and as a result the weight loss recorded is not only due to loss of moisture.
Highly unsaturated oils can oxidize d~~g the evaporation process and an increase in . . .
weight may actually be recorded, while triglycerides may undergo hydrolysis by the
moi sture initially present giving rise to FF As.
The distillation approach is based on the azeotropic property ofwater which is the
ability of water to form constant boiling mixtures with many organic solvents in which it
is immiscible when cold. The AOCS distillation method (Ca 2a-45) involves mixing 20-
200 g of the test sample with 100-300 ml oftoluene. The mixture is heated and water is
distilled out of the sample as a constant boiling mixture with the solvent. The distillate
separates with water forming the lower phase, whereupon its volume is measured in a
specially graduated tube. Although it is an accurate procedure per se, the distillation
method has many drawbacks that limit its use, such as its limited sensitivity (not
18
applicable to samples containing less than 0.5% moisture), the large sample size required,
and its lack of suitability for high-volume testing.
The "best" method available for moisture determination is the Karl Fischer
titrimetric method (e.g., AOCS Ca 2e-84), and it is the gold standard against which
proposed new methods for moi sture analysis are compared. As such, this method will be
discussed in detail, and proposed modifications that have been reported in the literature
will be surveyed to provide a c1ear picture of its current status.
2.2.2. Karl Fischer Method
The Karl Fischer (KF) method involves the chemical reaction of water with the
Karl Fischer reagent (mixture of iodine, sulfur dioxide, pyridine (or similar base) and
methanol), which consumes water stoichiometrically. The general reaction is as follows:
S02 + CH30H + B ( ) CH3S03 (HB)
12 ,1;+ H 20 + CH3S03(HB) + 2B~CH3S04(HB) + 2(HB)1 [2.1]
where B is a base, such as pyridine or imidazole. The general protocol involves mixing
5-25 g of oïl with anhydrous methanol (or 1-propanol) and titrating with Karl Fischer
reagent to a cherry-red colored end point. Due to the high error associated with manual
titration (0.6% relative error) (51), the,titratiop is frequently carried out automatically in a 1 i ,
"
c10sed vessel where the end point is determined potentiometrically. The automated
method can determine as low as 200 ppm of water. Although detection of the end point
by potentiometric measurement is more sensitive than in the case of manual titration, it is
still a titrimetric method and, as such, standardized reagents are required to obtain
consistent and accurate results. The need for standardization is eliminated by using
coulometry; an additional advantage of coulometric measurements is their higher
sensitivity, which in the case of the KF reaction extended the limit of detection to coyer
the range 1-25,000 ppm water.
Although the KF reagent is specific to water, the presence of aldehydes and
ketones can complicate the analysis by introducing side reactions. Two side reactions are
19
weIl known to take place during KF measurement which directly interfere with accurate
moi sture determination and are believed to occur "simultaneously" in the presence of
aldehydes/ketones. The first reaction is acetal/ketal formation where methanol reacts with
aldehydes/ketones to produce water molecules as shown in Eq. 2.2.
H H, /OCH 3
\=0 + 2 CH 30H .. + H 20 C
RI R/ "aCH 3
R R, /OCH 3
)=0 + 2 CH 30H .. + H 20 C
R/ "aCH 3 [2.2] R
The occurrence of such reactions increases the number of water molecules in the
system, leading to an overestimation of moi sture content in the sample being analyzed.
The second reaction is called bisulfite addition. Bislfite ions are common products in KF
systems, being formed reversibly by the reaction between water molecules and sulfur
dioxide under basic conditions (Eq. 2.3). In the presence of aldehydes/ketones, the bislfite
ions undergo addition reactions and form bisulfite-addition products as shown in the
reaction scheme below:
H \ e ® c=o + HS0 3 + HB
1 R
\=0 + soie + 2 HB®
1 [2.3]
20
As can be noted from these reaction equations, these reversible reactions consume
water, which is then released back into the system owing to the shifting of the
equilibrium to the left as water reacts with the KF reagent. Accordingly, these reactions
delay reaching of the end point, leading to an underestimation of the water content in the
sample when these reactions are not taken into account in setting the instrument
parameters.
The reaction conditions and rates of the above reactions were extensively studied
by many authors in order to minimize the interference as much as possible. In this regard,
Scholz (52) reported after studying the behavior of 44 difÎerent aldehydes and ketones
with different KF formulations that acetal/ketal formation reactions are acid-catalyzed
and that higher pH had a significant effect in reducing the extent of these side reactions.
He suggested use of the stronger base imidazole as a substitute for pyridine in the KF
. formulation in order to make the reaction conditions more basic. Scholz (52) agreed with
other authors who found that replacement of methanol with other solvents, such as 2-
methoxyethanol (53), dimethylformamide (54), and propylene carbonate (55), suppressed
acetal/ketal formation reactions, but he found that the bisulfite addition reaction is more
favored in nonalcoholic solvents; consequently, he suggested use of alcoholic solvents
and higher pH rather than use of nonalcoholic solvents as a means of suppressing side
reactions of aldehydes and ketones with KF reagents. Examination of different types of
alcoholic solvents (primary, secondary and tertiary alcohols) as an alternative to methanol
led Scholz to strongly recommend that f7cl1loroethanol and 2,2,2-trifluoroethanol be used
in the KF formulation instead of methanol. He reported that KF formulations using either
of these solvents with imidazole as the base had outstanding performance in
determination of water in the presence of aldehydes and ketones; in the same year he
patented a new KF reagent using those two solvents and imidazole (56). In 1987,
Andersson and Cedergren (57) reported that the bisulfite addition reaction is a slow
reaction and, therefore, suggested use of fast-reacting KF reagents or higher iodine
concentration to reduce the analysis time in order to make the onset of the bisulfite
addition reaction negligible. In addition, the authors (57) found no observable
improvements when 2,2,2-trifluoroethanol was used in place of 2-methoxyethanol. Bizot
(58) and Cedergren (59) reported that addition of sorne chemicals such as formamide to
21
pyridine-buffered methanolic KF reagent was found to speed up the KF reaction by a
factor of 100, which in turn can minimize the problems connected with slow reaction
rates between KF reagents and water. Altematively, Scholz (60) reported that
replacement of pyridine with imidazole (more basic than pyridine) was found to increase
the speed of the reaction of the KF reagent with water and ultimately minimize the effect
of both the bisulfite addition reaction and acetal/ketal formation. In 1994, Oradd and
Cedergren (61) showed that in order for imidazole to speed up the reaction the pH has to
be in the range 7-10.
Besides the side reactions of water productionlconsumption due to presence of
active carbonyl groups, there are other important side reactions associated with iodine. In
coulometric KF analysis, iodine is formed at the anode and reacts quantitatively with
water present in the system as shown in Eq. 2.1, with the end point of the titration being
the point at which iodine is no longer consumed. Iodine is known to be a strong oxidizing
agent and can react with substances such as thiosulfite, thiosulfate, ascorbic aeid,
hydrazines, hydroxylamines, Tl, Sn2+, In+ and Cu+, etc., in the matrix being analyzed,
resulting in overestimation of the moi sture content. On the other hand, certain oxidizing
agents such as Cu2+, Fe3+, N02", Br2, Ch and quinines would oxidize r (one of the KF
reaction products) to h, and thus the presence of these species in the matrix would lead to
underestimation of the moi sture content. In addition, h can combine with r to pro duce
the triiodide ion (h):
l , . i 1°'" 1" 2;h-,~ 3 [2.4]
Although this ion will eventually react with water, it has been reported that reactions
involving the triiodide ion are about 4-fold slower than those of iodine itself (62), which
may lead to underestimation of water content. To minimize the above-mentioned
problems, it is strongly suggested to use fast-reacting KF reagents combined with
optimum conditions in order to speed up the reaction so that less side reactions will have
chance to occur (63,64).
Another important modification of the KF method was that introduced to make it
suitable for the determination of water in hydrophobic liquids such as edible oils and
lubricants, because the KF reagents as originally formulated (65) were not miscible with
22
such samples. It was suggested to modify the polarity of the KF working medium by
adding solvents such as chloroform (66), xylene, hexanol, I-propanol, l-octanol (67),
propylene glycol and others. Partial solubilization of oil samples in KF reagents was
found to be one of the main reasons for the commonly observed variation in results
obtained by the KF method (68).
Most of the progress that has been made in understanding and improving KF
reactions has been adopted commercially. An important consideration in making
commercial formulations is the toxicity of the reagents, and the pyridine employed in the
original KF reagent has largely been replaced by the less noxious imidazole, which also
serves to speed up the KF reaction, as discussed above. Different formulations of KF
reagents are available under various brand names, the most common being
HYDRANAL ®, representing pyridine-free reagents, which are produced in different
formulations to suit a variety of applications. On the other hand, this complicates the
univers al comparison of KF results as the results obtained depend on the formulation
used.
KF reaction cell. In conjunction with the work done with reagent formulations to
make the KF method faster and less susceptible to side reactions, substantial development
has taken place in relation to the design of the reaction vesse!. The conventional
coulometric cell inc1udes a potentiometric detector, a cathodic and an anodic
compartment, and a diaphragm. The cathode is in electrolytic contact with the anode
through the diaphragril, whi~h pl~ysa:v:éry·important role in restricting the iodine that is
being produced at the anode from reaching the cathode, where it can be reduced by
reaction products such as thiosulfate, hydrogen sulfide and other products that are
believed to be produced at the cathode (69,70). Optimization of the dimensions of the
cathode and anode and the possibility of removing the diaphragm have been extensively
studied to make the design simpler and facilitate maintenance. As a result, many designs
have been proposed with various cathode dimensions, with or without a diaphragm, and
with the use of constant or pulsed currents (71-79). The use of a diaphragm eliminates the
possibility of iodine being reduced by products at the cathode and allows the use of
electrolytes other than the KF reagent in the cathodic compartment of the cell. Clogging
of the diaphragm, however, is a common problem and the cell requires long conditioning
23
times before startup (~2.5 hl. On the other hand, diaphragm-free systems have shorter
conditioning times and are easier to maintain, but they suffer from the formation of
oxidizable reduction products at the cathode. This problem has been minimized
considerably by implementing rapidly reacting reagents combined with optimizing the
dimensions of the cathode and the current density. The latter were found to be critical
factors in determining the analytical accuracy obtained with the coulometric cell (72, 77-
80). The type of current used in the coulometric ceIl, pulsed or constant, was found to be
critical, especially for diaphragm-free cells, in reducing the rate of development of
oxidizable products at the cathode and also to minimize the drift at the end point. In this
regard, pulsed current was found to be the most suitable for diaphragm-free cells (78-80).
Investigation of such cells also indicated that the titration rate is a critical parameter in
relation to analytical accuracy (79), but the optimum varies from one system to another.
KF detectors. Many end point detectors have been used with the KF method.
Determining which detector is best from the literature is not easy because comparisons
are not straightforward as the end-point detection is influenced by a number of factors.
These include the mode of titration (volumetric or coulometric), cell design, electrode
response, the extent of background drift caused by moisture diffusion, side reactions and
the use of different reagents and additives in the titration.
Early on, amperometric detectors were commonly in use, largely supplanted later
(81,82) by controlled-current potentiometers (bipoteniometry) (59, 83-86). In general,
both these types of detectol'S use a' consùmt current between two platinum electrodes
while measuring the voltage needed to sustain the current. At the equivalence point of the
titration, there is a sharp drop in the voltage required to sustain the current because excess
12 is present and the current can be carried at a very low voltage in its presence. Although
controlled-current potentiometers have better accuracy and shorter analysis times than
amperometric detectors, their accuracy was found to be affected by both the electrode
dimensions and the external current generated between the cathode and the anode.
Recently, zero-current potentiometers (true potentiometry) have received substantial
interest (66,71,87,88) because they offer better selectivity and do not suffer from the
electric field interference generated in the coulometric cell. Although it is the detector of
choice in modem KF coulometric ceIls, this type of detector must meet certain
24
requirements to perform well such as a suitable reference electrode system and a
calibration procedure to establish the relationship between the redox potential and the
concentration of excess iodine (87).
Spectrophotometric end-point detection. Upon addition of the KF reagent to a
sample, the original dark brown color due to the presence of iodine changes toward light
yellow depending on the amount of water in the sample that has reacted with the KF
reagent. This change in color implies that the development of a colorimetric method for
the KF method would be possible; however, very little work has been done in this regard.
The earliest attempt to employ colorimetric detection in KF titration involved measuring
the absorbance of the sample at 525 nm (89). This approach allowed detection of 100
ppm water with a precision of ±3 ppm. Subsequently, Dahms (90) patented a colorimetric
method for determination of moi sture using the KF reaction based on the injection of the
sample into a fixed volume of KF reagent and colorimetric measurement of the change in
optical density (OD) at 520 nm. The amount ofwater in thesample was determined from
a calibration curve established by measuring the OD of the reagent mixed with defined
amounts of water. Although the author used 520 nm as a working wavelength, he noted
that the KF reagent absorbs over a wide spectral range (500-620 nm) and the sensitivity
of the method depended on the wavelength used. This approach was adapted for use with
a flow injection analysis (FIA) system (91) and improved to correct for factors affecting
OD measurements such as dilution effects, the transmission characteristics of the cuvette,
the refractive index of the sample, e~co'" through the use of a reference dye that has an "Ii f
absorption (600 nm) far from the working wavelength (set at 420 nm) (92). A system
based on this approach combined with specially designed ready-to-use sealed vials
containing the KF reagent was developed (93,94). Although this system appears simple to
use, it has not been independently validated.
Flow injection analysis. Due to the many interferences affecting KF titrimetric
methods, various approaches have been explored to carry out KF analysis. As indicated
earlier, many authors suggested speeding up the KF reaction as a means to both overcome
the problems of side reactions and increase accuracy by minimizing the problems
inherent to coulometric analysis. These issues become even more important when one
considers FIA as a means of automating KF moi sture determinations. In 1980, Kagevall
25
et al. (91) developed an FIA system for determination of moi sture in organic solvents
using both potentiometric and spectrophotometric detectors. The method used a fast
reacting one-component KF reagent (formamide/pyridine buffer instead of pyridine
alone) which was diluted with methanol as a carrier solvent. The system was capable of
analyzing up to 120 samples/h and covered a range of 100-50,000 ppm moisture. The
authors noted that the potentiometric detector gave a better SD (:::;0.5%) than that
obtained spectrophotometrically. The efficacy of the system in minimizing side reactions
in iodine-consuming samples was tested using penicillin as the sample (95), with the
results indicating a considerable decrease in side reactions. A similar, but improved FIA
system was introduced by Escott and Taylor (96) to analyze gasoline-alcohol blends. A
methanol-xylene mixture was used to dilute samples and make them miscible with the
methanolic KF reagent. Quantification was based on direct potentiometric measurement
and the system, which was equipped with an autosampler, was able to analyze up to 60
samples/h over a range of 1-1500 ppm moisture. The results tended to be higher than
those obtained using a titrator, but the reproducibility was better. The variables affecting
the performance of the KF method in the FIA mode using potentiometric or
spectrophotometric detectors have been thoroughly investigated by Dantan et al. (97).
The authors suggested new FIA designs for both types of detectors to pro vide more
flexibility in reagent concentrations as well as on-line analysis capability. The optimized
methods permit rapid, precise and automated determinations of moi sture in a wide range
of samples over the range of 100-50,000 ppm with a reproducibility of better than 3% • l, j
RSD. According to the authors, alteraÙon of~ertain parameters, such as injection volume,
detection wavelength and/or the concentration of KF reagent, could further extend the
analytical range, with the presence of traces of water in the carrier solvent being the main
factor determining the detection limits of these methods.
Current status of the KF method. As can be seen from the above discussion, most
of the modifications of the KF method proposed to date have focused on minimizing the
inherent problems associated with the KF reaction itself. These modifications have
minimized the problems of side reactions and increased the diversity of samples that can
be analyzed, but at the cost of complicating the method due to the introduction of
different KF reagents and apparatus designs. The modifications that were introduced over
26
the years made the method suitable for the analysis of almost an types of samples;
however, ultimately the accuracy of the method is dependent on selecting the right
combinations of a long list of variables and operating parameters. The selection of
variables is difficult because it requires not only knowing what is in the sample that is
being analyzed but also understanding clearly the chemistry involved. In this regard,
Margolis (68,98) noted that systematic biases were observed in interlaboratory
collaborative studies and depended on both the type of apparatus and the nature of the
solvent system employed in the measurement. He tested a number of hydrocarbon-based
lubricating oils using different KF methods (volumetric and coulometric) and different
analysis conditions (different solvent compositions and different instrumental parameters)
and concluded that the measurements were very much dependent on the nature and
concentration of the non-polar organic solvent and the miscibility of the oil with the KF
reagent. The severity of the bias can be minimized by selecting the appropriate working
parameters and reagent composition for the sample at hand, but these are not readily
evident in many cases. Recently, Larsson and Cedergren (99) studied factors influencing
the accuracy and precision of moisture determination in oil using a diaphragm-free KF
coulometric cell using eight different types of commercial coulometric reagents and
various modifications. The best results were obtained with a homemade coulometric cell
with fully adjustable parameters and reagent formulations not commercially available.
Although one of the authors (Cedergren) has spent more than 20 years working with the
KF method, he declared that there is more research needed to better understand the
function of the cathode reaction in the diaphragm-free coulometric cell when using
different types of KF media. Aside from these issues, the coulometric cell has the
disadvantage of requiring regular maintenance (weekly) and long conditioning times.
Even with its many problems, the KF method remains the gold standard for moi sture
determination as no better alternative has been developed.
2.2.3. Infrared Methods
At first glance, FTIR spectroscopy would appear to be a viable technique for the
measurement of moisture in oils, as water absorbs strongly in the IR portion of the
spectrum (Figure 2.7), exhibiting intense and readily measurable bands in the regions of
27
3700-3000 and 1670-1630 cm- l due to its O-H stretching and H-O-H bending vibrations,
respectively (100).
0.4
0.3
~ CJ
= c: 0.2 ,.Q "-0 C7.l
,.Q
< 0.1
o.o~---
4000 3500 3000 2500 2000 1500 1000
Wavenuniber (cm-1)
Figure 2.7. FTIR spectrum ofpure water.
However, FTIR moisture analysis in oils is complicated by spectral interferences
from other OH-containing constituents, such as FF A, mono- and diglycerides, alcohols,
and hydroperoxides, which may be present in substantial amounts (in terms of their
impact on moisture quantification both in crude oils and in refined oils that have
undergone auto-oxidation to any significant extent). Quantification of moisture is further
confounded by the hydrogen bonding interactions between water and the se species as
, weIl as between water and carbonyl-containing secondary oxidation products (aldehydes
and ketones), as such interactions affect the positions, band shapes, and intensities of the
water absorptions. Chemometric techniques such as PLS can, in principle, be employed
to compensate for these types of "matrix effects" but, in practice, their utility is limited by
the difficulty of meeting the criterion that aIl possible sources of spectral variability must
28
be represented within the training set used to develop a calibration model. This limitation
is well illustrated by previous work conducted by the McGill IR Group on the
determination ofhydroperoxides in edible oils based on their Q-H stretching band at 3444
cm-l, which is subject to effectively the same confounding effects as moi sture analysis
(101,102). Recently, Che Man and Mirghani (50) published a PLS-based FTIR method
for the determination of moisture in crude palm oïl based on a set of calibration standards
prepared by spiking oïls with water to coyer a range of 0-13% moi sture content. The
method is simple to implement, involving recording the spectra of neat oil samples in a
1 OO-~m transmission cell and using the PLS algorithm to predict the moi sture content.
However, validation of this method by predicting an independent test set was not
performed, and hence there is no evidence that unknowns would be modeled adequately
by the limited number of spiked oïl samples used to develop the PLS calibration. Indeed,
the physical stability of the calibration standards, and hence the validity of the calibration
model, is questionable since phase separation of oïl standards having moisture contents as
high as 13% would seem likely. Apart from these issues, this work does not serve as the
basis for a global method, as the calibration model developed is specific for crude palm
oïl and does not take into account the complicating factors outlined above that must be
considered in the development of a generalized FTIR method for the determination of
moisture in edible oils.
A completely different approach to overcoming matrix effects was taken by van
de Voort et al. (103) for the FTIR dete~ination of moi sture in new and used mineral
based lubricants, which are quite different matrices from edible oils but present similar
challenges from an FTIR analytical perspective. The approach invoives splitting the oïl
sample into two halves and treating one half with a dimethoxypropane (DMP)/dioxane
mixture to convert water present in the oïl into acetone via the reaction in Eq.2.5 prior to
recording its FTIR spectrum.
OCH3
1 CH3-C-CH3
1 OCH3
2,2-dimethoxypropane
® H
[2.5]
29
The other half of the sample is simply diluted with dioxane, and its spectrum is
subtracted from that of the DMP-treated portion to generate a differential spectrum. The
amount of acetone that is produced in the reaction is quantified by measuring its v (C=O)
absorption at 1740 cm-1 in the differential spectrum and is related to the amount of water
present in the initial sample using a simple calibration equation. Owing to the high molar
absorptivity of this acetone band, very good sensitivity (±50 ppm water) was obtained
over the range 0-2000 ppm. Thus, the use of a reagent that reacts stoichiometrically with
water to produce a strongly IR-absorbing product followed by the application of
differential spectroscopy to isolate the spectral features associated with this conversion
provided a generalized method for the determination of moi sture in mineral-based
lubricants, without the complexities inherent in chemometric modeling of the sample
matrix. This approach, however, is not directly applicable to edible oils because the
triacylglycerol ester linkages give rise to an intense v (C=O) absorption that would
completely mask the v (C=O) band of acetone in a transmission-cell spectrum, except at
extremely short path lengths. Moreover, the use of ATR, which inherently provides a
short effective path length, to overcome this limitation is impractical owing to the
difficulty of preventing the sample from picking up moi sture from the environment or
losing the acetone produced, nor would it provide sufficient sensitivity for the analysis of
moisture at the low levels present in edible oils.
With regard to moisture analysis in edible oils by vibrational spectroscopic
techniques other than FTIR spectroscopy, a few papers on NIR methods have appeared in
the literature. The first paper, published in 1970 by Vornheder and Brabbs (104),
extended work published by other authors who determined water content in a variety of
samples (organic solvents (105) and food materials (106,107», but not edible oils. The
method involves mixing. oil samples with a polar solvent (DMSO) in order to extract
water molecules from the oil. The intensity of the water absorption band in the NIR
spectrum of the solvent is then measured in the range 2.1-1.8 J.lm (max absorption at 1.94
J.lm (5160 cm-1), and the water content is determined using a simple Beer's law equation.
Although the working range of the method depends on the sample:solvent ratio, the limit
of detection of the method was 300 ppm. Recently, Cozzolino et al. (46) used PLS to
30
~"
establish a method for moi sture analysis in fish oil from the NIR spectra of neat oil
samples. The calibration equation had R2 of 0.8.
2.3. CONCLUSION
The standardized methods for the determination of both FF A and moi sture in
edible oils established in the middle of the twentieth century continue to be the basis for
the determination of these crucial oil quality parameters. Many modifications of these
methods or alternative techniques have been proposed, but it has been shown in the
discussion above that none of these have fully addressed the need for rapid, accurate and
simple instrumental methods that would be suited to automated process and quality
control. FTIR spectroscopy has the potential to meet this requirement, but this promise
has not yet been fulfilled in fairly extensive work to date on FF A analysis or in a single
study on moi sture analysis. Accordingly, the research described in the following chapters
of this thesis was undertaken to overcome the limitations of the FTIR methods developed
to date and allow the potential ofFTIR analysis to be realized.
2.4. REFERENCES
1. O'Brien, R. D., Fats and ails: Formulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998.
2. Official Methods and Recommended Practices of the American ail Chemists 1
Society, 4th edn., Champaign, IL, 1989, AOCS Official Method Ca-5a-40.
3. Determination of the Free Fatty Acids in European Commission Regulation (EED) No. 2568/91, Official J01;1I'11al of the European Communities, No. L 248 (5.9.91), p. 6. " ,
4. Ayers, C. W. Estimation of the Higher Fatty Aeids C7-C18. Anal. Chim. Acta, 15, 77-83 (1956).
5. Baker, D. A Colorimetrie Method for Determining Free Fatty Acids in Vegetable Oils., J. Am. ail Chem. Soc., 41, 21-23 (1964).
6. Lowry, R. R. and Tinsley, 1. J. Rapid Colorimetrie Determination of Free Fatty Acids, J. Am. ail Chem. Soc., 53, 470-473 (1976).
7. Bains, G. S., Rao, S. V. and Bhatina, D. S. A Colorimetrie Procedure for the Estimation of fat Acidity in Peanuts and Peanut Meals. J. Am. ail Chem. Soc., 41, 831-832 (1964).
8. Ekstrom, L-G. An Automated Method for Determination of Free Fatty Acids. J. Am. ail Chem. Soc., 58, 835-838 (1981).
31
9. Kwon, D. Y. and Rhee, 1. S. A Simple and Rapid Colorimetric Method for Determination of Free Fatty Acids for Lipase Assay. J. Am. Oil Chem. Soc., 63, 89-92 (1986).
10. Brenardez, M., Pastoriza, L., Sampedro, G., Herrera, J. J. and Cabo, M. L. Modified Methods for the Analysis of Free Fatty Acids in Fish. J. Agric. Food Chem., 53, 1903-1906 (2003).
Il. Canham, 1. S. and Pacey, G. E. Automated Free Fatty Acids Determination Using Flow Injection Analysis Solvent Extractions. J. Am. Oil Chem. Soc., 64, 1004-1007 (1987).
12. Puchades, R, Suescun, A. and Maquieira, A. Determination of Free Fatty Acids in Foods by Flow Injection. J. Sei. Food Agric., 66, 473-478 (1994).
13. Zhi, Z-L, Rios, A. and Valcârcel, M. An automated Flow ReversaI InjectionlLiquid-Liquid Extraction Approach to the Direct Determination of Total Free Fatty Acids in Olive Oils. Anal. Chim. Acta, 318, 187-194 (1996).
14. Linares, P., Luque de castro, M. D. and Valcârcel, M. Direct Automatic Determination of Free Acidity in Oils Flow-Injection Analysis. Anal. Chim. Acta, 255,431-436 (1989).
15. Nouros, P. G., Constantino s, A. G. and Polissiou, M. G. Automated Flow Injection Spectrophotometric Non-Aqueous Titrimetric Determination of the Free Fatty Acid Content of Olive Oil. Anal. Chim. Acta, 351, 291-297 (1997).
16. Mariotti, E. and Mascini, M. Determination of Extra Virgin Olive Oil Acidity by FIA-Titration. Food Chem., 73,235-238 (2001).
17. Walde, P. A Colorimetric Determination ofFatty Acids as a New Assay of Lipase in Reverse Micelles. J. Am. Oil Chem. Soc., 67, 110-115 (1990).
18. Walde, P. and Nastruzzi, C. Application of a New, Simple and Economic Colorimetric Method for the Determination of Non-Esterified Fatty Acids in Vegetable Oils. Food Chem., 39, 249-256 (1991).
• 1, 1
19. Takamura, K. and Hayakaw~: Y. betermination of Traces of Free Acid by the Polarographic Reduction of Quinones. Microchem. J., 17,546-550. (1972).
20. Kusu, F., Fuse, T. and Takamura, K. Voltammetric Determination of Acid Values of Fats and Oils. J. AOAC Int., 77, 1686-1689 (1994).
2l. Fuse, T., Kusu, F. and Takamura, K. Voltammetric Determination of Free Fatty Acid Content in Fats and Oils. Bunseki Kagaku, 44, 29-33 (1995).
22. Fuse, T., Kusu, F. and Takamura, K. Determination of Higher Fatty acids in ails by high-Performance liquid Chromatography with Electrochemical detection. J. Chromatogr., 764, 177-182 (1997).
23. Fuse, T., Kusu, F. and Takamura, K. Determination of Acid Values in Fats and ails by Flow Injection Analysis with Electrochemical Detection. J. Pharm. Biomed. Anal., 15, 1515-1519 (1997).
32
~--
24. Takamura, K., Fuse, T., Arai, K. and Kusu, F. A Review ofa New Voltammetric Method for Determining Acids. J. Electroanal. Chem., 468,53-63 (1999).
25. Kotani, A., Kusu, F. and Takamura, K. New Elecrochemical Detection Method in High-performance Liquid Chromatography for Determining Free Fatty Acids. Anal. Chim. Acta, 465, 199-206 (2002).
26. Lapshina, T. M., Tur'yan, Y. I. and Danilchuk, S. I. pH-Metric Determination of Acid Numbers for Oils. J. Anal. Chem. USSR, 46, 1150-1158 (1991).
27. Tur'yan, Ya. I., Berezin, O. Yu., Kuselman, I. and Shenhar, A. pH-Metric Determination of Acid Values in Vegetable Oîls without Titration. J. Am. Oil Chem. Soc., 73, 295-301 (1996).
28. Tur'yan, Ya. I., Strochkova, E., Berezin, O. Yu., Kuselman, I. and Shenhar, A. pH-Metric Determination of Acid Numbers in Petroleum without Titration. Talanta, 47,53-58 (1998).
29. Kuselman, I,.Tur'yan, Ya. I., Burenko, I. and Anisimov, B. pH-Metric Determination of Acid Values in Oilseeds without Titration. Talanta, 49, 629-637 (1999).
30.
31.
32.
33.
34 ..
35.
36.
37.
38.
39.
40.
Strochkova, E. Tur'yan, Ya. I. and Kuselman, I. pH-Metric Determination for Acid Number in Hydraulic Oils without Titration. Talanta, 50, 1135-1139 (1999).
Lingane, J. J. Automatic Potentiometric Titrations. Anal. Chem., 20, 285-292 (1948).
Muller, R H. Electronic Trigger Circuit for Automatic Potentiometric and Photometrie Titrations. Anal. Chem., 20, 795-797 (1948).
Shaffer, P. A., Briglio, A. and Brockman, A. J. Instrument for Automatic Continuous Titration. Anal. Chem., 20, 1008-1014 (1948).
Delahay, P. Electrical DifferentiaI Method of Measurement in Electrotitrations: Application to Potentiometric Titrations. Anal. Chem., 20, 1212-1215 (1948).
Furman, N. H. Potentiometric Titrations. Anal. Chem., 22, 33-41 (1950).
http://www.mt.comlmtlresourcedetail/appEdStyle.jsp?m=t&key=I2Mjg4NjM1Mj
Vaughan, G. A. and Swithenbank, J. 1., Acetone as Solvent in the Enthalpimetric Titration of Acidic Substances. Analyst, 90, 594-599 (1965).
Thomas, K. S. Analysis of FF A in Edible Oils by Catalyzed End-Point Thermometric Titrimetry (CETT). J. Am. Oil Chem. Soc., 80,21-24 (2003).
Lanser, A. c., List, G. R, Holloway, R K. and Mounts, T. L. FTIR Estimation of Free Fatty Acid Content in Crude Oils Extracted from Damaged Soybeans, J. Am. Oil Chem. Soc., 68, 448-449 (1991).
Ismail, A. A., van de Voort, F. R and Sedman, J. Rapid Quantitative Determination of Free Fatty Acids in Fats and Oils by FTIR Spectroscopy. J. Am. Oil Chem. Soc., 70, 335-341 (1993).
33
41. Man, Y. B., Moh, M. H. and van de Voort, F. R Determination of Free Fatty Acids in Crude Palm Oil and Refined-Bleached-Deodorized Palm Olein Using Fourier Transform Infrared Spectroscopy, J. Am. ail Chem. Soc., 76, 485-490 (1999).
42. Bertran, E., Blanco, M., Coello, J., Iturriaga, H., Maspoch, S. and Montoliu, 1. Determination of Olive Oil Free Fatty Acid by Fourier Transform Infrared Spectroscopy, J. Am. ail Chem. Soc., 76, 611-616 (1999).
43. Verleyen, T., Verhe, R, Cano, A, Huyghebaert, A and De Greyt, W. Influence of Triacylglycerol Characteristics on the Determination of Free Fatty Acids in Vegetable Oils by Fourier Transform Infrared Spectroscopy, J. Am. ail Chem. Soc., 78,981-984 (2001).
44. Caiiada, M. J. A., Medina, A R and Lendl, B. Determination of Free Fatty Acids in Edible Oils by Continuous-Flow Analysis with FT-IR Spectroscopic Detection, Appl. Spectrosc., 55, 356-360 (2001).
45. Zhang, H-Z and Lee, T-C. Rapid Near-Infrared Spectroscopic Method for the Determination of Fatty Acid in Fish and Its Application in Fish Quality Assessment. J. Agric. Food Chem., 45, 3515-3521 (1997).
46. Cozzolino, D., Murray, 1., Chree, A. and Scaife, J. R Multivariate Determination of Free Fatty Acids and Moisture in Fish Oils by Partial Least-Ssquares Regression and Near-Infrared Spectroscopy. LWT Food Sei. Technol., 38, 821-828 (2005).
47. Man, Y. B. and Moh, M. H. Determination of Free Fatty Acids in Palm Oil by Near-Infrared Reflectance Spectroscopy. J. Am. ail Chem. Soc., 75, 557-562 (1998).
48. Muik, B., Lendl, B., M-D, A and Caiiada, M. J. Direct, Reagent-Free Determination of Free Fatty Acid Content in Olive Oil and Olives by Fourier Transform Raman Spectrometry. Anal. Chim. Acta, 487, 211-220 (2003).
49. Mailer, R J.' Rapid Evalua~ion, of Olive Oil Quality by NIR Reflectance Spectroscopy. J. Am. Oil Chem. 'Soc., 81, 823-827 (2004).
50. Man, Y. B. C and Mirghani, M. E. S., Rapid Method for Determining Moisture Content in Crude PaIm Oil by Fourier Transform Infrared Spectroscopy. J. Am. ail Chem. Soc., 77, 631-637 (2000).
51. Official Methods and Recommended Practices of the American ail Chemists' Society (AOCS, Champaign, IL, 1989), 4th edn.
52. Scholz, E. Karl Fischer Titrations of Aldehydes and Ketones. Anal. Chem., 57, 2965-2971 (1985).
53. Fischer, F. and Schiene, R Dead-Stop Determination of Traces of Water in Aliphatic Ketones with Karl Fischer Solution. Zeitschrift fuer Chemie, 4 (2), 69-70 (1964).
34
54. Klimova, V. A., Shennan, F. B. and L'vov, A. M. Micro-Detennination ofWater by Fischer Reagent Prepared with N,N'-Dimethylfonnamide Instead of Methanol. Izvestiya Akademii Nauk SSSR, Seriya Khimicheskaya, 12, 2599-2602 (1967).
55. Mitsubishi Chemical Industries, Gennan Patent Application DE 30 40 474 (1980).
56. Hydranal Composite 5K, leaflet from Riedel de Haen AG, D-3016 seeize.
57. Nordin-Andersson, 1. and Cedergren, A. Coulometric Detennination of Trace Water in Active Carbonyl Compounds Using Modified Karl Fischer Reagents. Anal. Chem., 59, 749-753 (1987).
58. Bizot, J. Automatic Method for Coulometric Detennination of Minute Quantities ofWater. Bull. Soc. Chim. France, 151- 157 (1967).
59. Cedergren, A. Reaction Rates Between Water and some Modified RapidlyReacting Karl Fischer Reagents. Talanta, 25,229-232 (1978).
60. Scholz, E. Karl-Fischer reagents without pyridine. 3. Accuracy of the Detennination ofWatèr. Fresenius Z. Anal. Chem., 306, 394-396 (1981).
61. Oradd, C. and Cedergren, A. Reaction Rates of Water in Imidazole-Buffered Methanolic Karl Fischer Reagents. Anal. Chem., 66, 2603-2607 (1994).
62. Verhoef, J. C. and Bahrendrecht, E. Mechanism and Reaction Rate of the KarlFischer Titration Reaction. Part 1. Potentiometric Measurements J. Electroanal. Chem., 71, 305-315 (1976).
63. Cedergren, A. and Oradd, C. Coulometric Study of the Influence of Aldehydes and Ketones on the Karl Fischer Titration of Water Using Standard Pyridine !Methanol Reagents. Anal. Chem., 66,2010-2016 (1994).
64. Oradd, C. and Cedergren, A. Coulometric Study of the Recovery Rates for Karl Fischer Titration of Water in Aldehydes and Ketones Using Rapidly Reacting Methanolic and 2-Methoxyethanolic Reagents. Anal. Chem., 67, 999-1004 (1995).
65. Fischer, K. A New Methoçl forth'e'Analytical Detennination of the Water Content of Liquids and Solids: Angew. Chem., 48, 394-396 (1935).
66. Cedergren, A. Coulometric Study of Reaction Rates of Water in Pyridine and Imidazole-Buffered Methanolic Karl Fischer Reagents Containing Chlorofonn. Anal. Chem., 69, 3682-5687 (1996).
67. Isengard, H-D and Kerwin, K. ProposaI of a New Reference Method for Detennining Water Content in Butter Oil. Food Chem., 82, 117-119 (2003).
68. Margolis, S. Amperometric Measurement of Moisture in Transfonner Oil using Karl Fischer Reagents. Anal. Chem., 67, 4239-4246 (1995).
69. Katoh, H. Fujimoto, Y. and Kuwata, Y. A Study of the Coulometric Cathode Reaction in the Karl Fischer Reaction. Anal. Sei. 7,299-302 (1991).
70. Scholz, E. Karl Fischer Coulometry - the Cathode Reaction. Fresenius J. Anal. Chem. 348,269-271 (1994).
35
71. Cedergren, A. and Jonsson, S. Diaphragm-Free Cell for Trace Determination of Water Based on the Karl Fischer Reaction using Continuous Coulometric Titration. Anal. Chem. 69,3100-3108 (1997).
72. Cedergren, A. and Jonsson, S. Progress in Karl Fischer Coulometry using Diaphragm-Free Cells. Anal. Chem. 73,5611-5615 (2001).
73. Dahms, H. High Current Cou1ometric KF Titrator. US Patent 5300207 (1994).
74. Nichugovskii, G. F. Cou1ometric Cell ofSimplified Design for Titration ofWater in a Fischer's Reagent Medium Zh. Anal. Khim. 32,2124-2126 (1977).
75. Katoh, H. Fujimoto, Y. and Kakuda, M. Effects of the Cathode Current Density in the Karl Fischer Coulometric Titration Method. Anal. Sei. 8,575-577 (1992).
76. Nichugovskii, G. F. and Yakolev, P. P. Coulometric Determination ofWater with Fischer's Reagent in a One-Chamber Cell. Zh. Anal. Khim., 33, 684-687 (1978).
77. Nordmark, U. and Cedergren, A. Conditions for Accurate Karl Fischer Cou1ometry using Diaphragm-Free Cells. Anal. Chem. 72, 172-179 (2000).
78. Nordmark, U., Rosvall, M. and Cedergren, A. Optimum Conditions for Pulse Generation in Daphram-Free Karl Fischer Coulometry. Fresenius J. Anal. Chem., 368, 456-460 (2000).
79. Nordmark, U. and Cedergren, A. Progress in Pulsed-Current Karl Fischer Coulometry using Diaphragm-Free Cells. Fresenius J. Anal. Chem., 367,519-524 (2000).
80. Borsten, H. and van Lamoen, F. L. J. Potentiometric Method for Karl Fischer Titration. Anal. Chem., 27, 1639-1640 (1955).
81. Brown, J. F. and Volume, W. F. An Automatic Fischer Titration Unit for Laboratory Use. Analyst, 81, 308-315 (1956).
82. Cedergren, A. Cou1ometric Trace Determination of Water by using Karl Fischer Regent and Potentiometric End-Point Detection. Talanta, 21, 367-375 (1974).
• j • I~ , 1
83. Lindbeck, M. R.and Freund,' H: Microdetermination of Water, using Rapid Controlled Potential Cou1ometry in a Karl Fischer System. Anal. Chem., 37, 1647-1650 (1965).
84. Cedergren, A. Comparison between Amperometric and True Potentiometric EndPoint Detection in the Determination of Water by the Karl Fischer Method. Talanta, 21, 553-563 (1972).
85. Harris, D. C. Quantitative Chemical Analysis, 5th edn., W. H. Freeman and Company, New York, 1999, pp. 498-499
86. Cedergren, A. Comparison between Bipotentiometric and True Potentiometric End-Point Detection using Rapidly Reacting Karl Fischer Reagents. Anal. Chem., 68,3679-3681 (1996).
87. Rosvall, M., Landmarl, L. and Cedergren, A. Computer-Controlled, Coulometric Karl Fischer Systems for Continuous Titration of Water Based on Zero Current Potentiometry. Anal. Chem., 70, 5332-5338 (1998).
36
88. Bakker, E. and Pretsch, E. Potentiometry at Trace Levels. Trends Anal. Chem., 20, 11-19 (2001).
89. Connors, K. A. and Higuchi, T. Spectrophotometric Detection of the End Point in the Karl Fischer Titration ofWater. Chemist-Analyst, 48, 91-92 (1959).
90. Dahms, H. Method and apparatus for colorimetrie analysis. US Patent 4005983 (1977).
91. Kâgevall, 1., Âstrom, O. and Cedergren, A. Determination of Water by FlowInjection Analysis with Karl Fischer Reagent. Anal. Chim. Acta, 114, 199-208 (1980).
92. Dahms, Harald. Method for colorimetricdetermination of water using Karl Fischer reagents and optical density measurements at two different wavelengths. US Patent 4786602 (1988).
93. Dahms, Harald. Improved Colorimetrie Karl Fischer Water Determination. PCT Int. Appl., WO 9221969 (1992).
94. Dahms, Harald. Sealed Vials Containing Improved Karl Fischer Solutions and Process for Water Determination using these Vials. US Patent 5179024 (1993).
95. Kâgevall, 1., Âstrom, O. and Cedergren, A. Minimization of Interference Effects from Iodine-Consuming Samples in the Determination of Water with the Karl Fischer Reagent in a Flow-Injection System. Anal. Chim. Acta, 132, 215-218 (1981).
96. Escott, R. E. and Taylor, A. F. Determination of Water by Flow Injection Analysis using Karl Fischer Reagent with Electrochemical Detection. Analyst, 110, 847- 849 (1985).
97. Dantan, N., Kroning, S., Frenzel, W. and Küppers, S. Comparison of Spectrophotometric and Potentiometric Detection for the Determination of Water using Karl Fischer Method under Flow Injection Analysis Conditions. Anal. Chim. Acta, 420, 133-142 (2000).
• , ,,' i il .1
98. Margolis, S. Effect ofHydrocarbon Composition on the Measurement ofWater in Oils by Coulometric and Volumetrie Karl Fischer Methods. Anal. Chem., 70, 4264-4270 (1998).
99. Larsson, W. and Cedergren, A. Coulometric Karl Fischer Titration of Trace Water in Diaphragm-Free Cells. Talanta, 65, 1349-1354 (2005).
100. Smith, B. Infrared Spectral Interpretation: A Systematic Approach, CRC Press, Boca Raton, FL, 1999, Chap. 1, p. 22.
101. van de Voort, F. R., Ismail, A. A., Sedman, J., Dubois, J. and Nicodemo, T. The Determination of Peroxide Value by Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 71, 815-826 (1993).
102. Ma, K., van De Voort, F. R., Sedman, J. and Ismail, A. A. Stoichiometric Determination of Hydroperoxides in Fats and Oils by Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 74, 897-906 (1997).
37
103. van de Voort, F. R., Sedman, J., Yaylayan, V., Saint Laurent, C. and Mucciardi, C. Quantitative Detennination of Moisture in Lubricants by Fourier Transfonn Infrared Spectroscopy. Appl. Spectrosc., 58, 193-198 (2004).
104. Vomheder, P. F. and Brabbs, W. J. Moisture Determination by Near-Infrared Spectrometry. Anal. Chem., 42, 1454-1456 (1970).
105. Keyworth, D. A. Detennination of Water by Near-Infrared Spectrophotometry. Talanta, 8, 461-469 (1961).
106. Rader, B. R. Determination of Moisture in Dried Vegetables and Spices. J. Assoc. Offic. Anal. Chem., 49, 726-730 (1966).
107. Gold, H. J. General Application of Near-infrared Moisture Analysis to Fruit and Vegetable Materials. Food Technol., 18(4), 184-185 (1964).
38
CHAPTER3
NEW FTIR METHOD FOR THE DETERMINATION OF FREE FATTY ACID IN
OILS
3.1. ABSTRACT
A rapid, practical and accurate FTIR method for the determination of free fatty
acids (FF A) in edible oils has been developed. Analogous to the AOCS titration
procedure, FTIR FF A determination is effected by an acid/base reaction but directly
measures the product formed rather than utilizing an endpoint based on an electrode
potential or color change. A suspension of a weak base, potassium phthalimide (K
phthal) in I-PrOH, is used to convert FFA present in oils to their carboxylate salts
without causing oil saponification, and differential spectroscopy is used to circumvent
matrix effects. Samples are tirst diluted with I-PrOH, then split, with one half treated
with the K-phthal reagent and the other half with I-PrOH (blank reagent), their spectra
collected, and a differential spectrum obtained to ratio out the invariant spectral
contributions from the oil sample. Quantitation of the percentage of FF A in the oil,
expressed as % oleic acid, based on measurement of the peak height of the v (COO")
absorption of the FF A salt formed yielded a calibration with a standard error of <0.020 %
FF A over the range of 0-4%. The method was validated by standard addition and the
analysis of Smalley check samples, the results indicating that the analytical performance
of the FTIR procedure is as good as or better than that of the standard titrimetric . "
procedure. As structured, the FTIR procedure is a primary method, as calibration is not
dependent on reference values provided by another method, and has performance criteria
which could lead to its consideration as an instrumental AOCS procedure for FF A
determination. The FTIR portion of the analysis is automatable, and a system capable of
analyzing ~60 samples/h has been developed that could be of benetit to laboratories that
carry out a large number of FF A analyses per day.
3.2. INTRODUCTION
FTIR spectroscopy is playing an increasingly important role in the analysis of
edible oils by providing simpler and more rapid techniques for determining common oil
quality parameters (1). FFA are common triacylglycerol (TAG) hydrolysis products in
39
crude oils and are fonned to sorne extent in refined oils as a result of oxidation or TAG
degradation during frying, impairing oil quality and functionality. Chemically, FFAs are
less stable than TAG and therefore more likely to oxidize and cause rancidity (2). The
standard method commonly used for FF A analysis is based on the titration of an oil
dissolved in alcohol with a strong base to a phenolphthalein endpoint (3,4). Although
simple, titrimetric methods are tedious, consume substantial amounts of solvent, and can
be problematic when dark crude oils are analyzed. The first application of FTIR
spectroscopy for FFA analysis was reported by Lanser et al. (5) in 1991. In this method,
which was developed for crude soybean oil, the FF A content was estimated from the FF A
v (C=O) band at 1710 cm- l. Owing to the overlap ofthis band with the very strong TAG
ester carbonyl absorption at 1746 cm-l, spectral deconvolution over the 2000-1600 cm- l
range was used to mathematically enhance spectral resolution. A calibration was derived
by spiking oleic acid into soybean oil at levels of 0.1-5% and yielded predictions of the
FF A content of soybean oils that matched the values obtained using the AOCS titrimetric
method to within ±0.5 percentage points. However, because the FTIR spectra were
acquired by simply placing each sample between two KBr windows, without the use of
an internaI standard, the accuracy of this method was limited by the resulting variability
in pathlength. In 1993, Ismail et al. (6) investigated two different FTIR approaches to the
quantitative detennination of FF A in edible oils. The first was based on measuring the
carboxylic acid v (C=O) band at 1711 cm-l in spectra acquired from oil samples applied
in their neat fonn ont9 an attenuated totaJ reflectance (ATR) crystal. Both calibration and , .
sample spectra were ratioed against the' ~pectrum of an FFA-free oi! of the same type as
the oi! being analyzed to reduce matrix effects. The second approach was an indirect
method based on the use of KOHlmethanol to extract the FF A present in the oi! and
convert them to their salts, followed by measurement of the v (COO") absorption band at
1570 cm- l in the spectrum of the methanol phase. This indirect method enhanced the
sensitivity of the analysis by concentrating the FF A in a small volume of methanol and
by utilizing an absorption band that is in a region free from major oïl spectral
Interferences. However, it had the disadvantage of requiring an additional procedural
step, and sorne saponification of oils by the KOHlmethanol reagent added was noted,
resulting in overestimation of the FF A content of the original sample. Verleyen et al. (7)
40
devised a more rigorous version of the method of Lanser et al. using peak height
measurements at 1711 cm-1 to develop workable calibrations for a variety of oils but
ultimately concluded that the calibrations were strongly oil dependent. Two other
publications have dealt with FF A analysis by FTIR spectroscopy, specifically in relation
to palm and olive oil, respectively (8,9), both using partialleast-squares regression (PLS)
to develop relationships between spectral changes and results obtained by standard
methods. The most sophisticated methodology is that of Cafiada et al. (10), who
developed an automated FTIR-based continuous-flow analysis system capable of
analyzing ~40 samples per hour using the indirect approach described by Ismail et al. (6).
However, even within the ~90 s required for analysis in this automated system, sorne
saponification of the oil can occur. In general, the main drawback associated with the
direct FTIR methods is their oil dependency, while the indirect FTIR methods are limited
by the possibility of errors due to saponification caused by the KOHImethanol reagent.
Hence, a simple, reliable, and robust FTIR method for FF A analysis is stilliacking.
The McGill IR Group recently developed an FTIR-based instrumental method to
replace the American Society for Testing and Materials (ASTM) titrimetric procedures
for the determination of acid number (AN) in mineraI and ester-based lubricating oils
(11). These ASTM methods are similar to those traditionally employed for the
determination of FF A content in edible oils except that they measure not only carboxylic
acids but also a variety of other acidic constituents, organic or inorganic, that accumulate
in lubricating oils either as a result, of Qxidation or as combustion by-products (12). " • 1
Because the FTIR AN method was specifically designed to meet the requirements of
lubricant analysis, it is not directly applicable to edible oils owing to their different
spectral characteristics. However, elements of this methodology have been adapted to
develop a new method for FF A determination that overcomes the limitations of both the
direct and indirect FTIR methods previously developed for FF A analysis. This paper
describes the princip les of this method as well as their practical implementation and
provides an evaluation of its performance by standard addition as weIl as by employing
AOCS SmaIley check samples.
41
3.3. EXPERIMENTAL PROCEDURE
Reagents and Standard Methods. AlI reagents used were of analytical grade.
Potassium phthalimide (K-phthal, 99+%) and hexanoic acid (99%) were obtained from
Aldrich (St. Louis, MO); I-propanol (l-PrOH) and iso-propanol were purchased from
Fisher Scientific Ltd. (Nepean, ON, Canada). AlI edible oils were obtained localIy and
samples were analyzed for FFA using AOCS method Ca 5a-40 (3). Mineral oil (C-171
polyalphaolefin) was obtained from Thermal-Lube (Montreal, QC, Canada) and used for
reagent blank determinations. A series of five oils pre-analyzed for FF A content were
obtained ,from the AOCS Smalley Check Sample program.
Instrumentation. The instrument used for this study was a Bornem WorklR
spectrometer (Bornem, Quebec, QC, Canada) equipped with a DTGS detector and purged
with dry air from a Balston dryer (Balston, Lexington, MA). Samples were analyzed by
aspirating them into a 500 flm Caf 2 transmission flow cell mounted on a sample shuttle
(Dwight Analytical, Toronto, ON, Canada). The spectrometer was controlled by an IBM
compatible Pentium 150-MHz PC running under proprietary Windows-based UMPlRE®
(Univers al Method Platform for InfraRed Evaluation) software (Thermal-Lube, Pointe
Claire, QC, Canada). AIl spectra were collected by co-adding 16 scans at a resolution of 8
cm-1 and a gain of 1.0.
Preparation of Calibration Standards. A series of eight standards covering the
range 0-4% FF A (expressed as perce~t.~g~ of oleic acid) were prepared by gravimetric 1 i f
addition of hexanoic acid to '3. retlned and deodorized soybean oil. The calibration curve
was obtained by linear regression of % oleic acid (%FF A) against the peak heights in the
FTIR spectra recorded for the standards by following the sample preparation and
analytical protocols described below.
Sam pIe Preparation for FTIR Analysis. Six grams of the oil sample was mixed
with 3 ml of I-PrOH in a 20-mL vial. Three-milli1iter aliquots of the diluted oïl were
placed in two centrifuge tubes, labeled BR (blank reagent) and RR (reactive reagent), to
which were added 7 mL of I-PrOH and 7 mL ofK-phtha1l1-PrOH (20 gIL), respectively.
AIl tubes were capped, shaken on a vortex mixer, and then centrifuged for a minimum of
5 min at 6000 rpm in a clinical centrifuge. It should be noted that K-phthal is virtually
42
insoluble in I-PrOH and was dispensed via bottle re-pipette ~s a fine dispersion which
was maintained by continuous and vigorous agitation on a magnetic stirrer.
Analytical Protocol. The transmission flow cell was loaded with ~2 mL of the BR
sample and its single-beam spectrum was recorded. After the cell was flushed with iso
propanol, -2 mL of the RR sample was loaded into the cell and its single-beam spectrum
was recorded and ratioed against that of its corresponding BR sample to pro duce a
differential spectrum. For quantitation of FFA, the peak height of the carboxylate v
(COO") band at 1570 cm-l in the differential spectrum was measured relative to a baseline
point at 2150 cm- l. A schematic diagram of the sample preparation procedure and
analysis is illustrated in Figure 3.1.
Diferential Spectrum
Figure 3.1. Schematic diagram illustrating the sample preparation procedure and the steps in the analytical protocolleading to the differential spectrum.
43
Validation. The FTIR method was vaHdated by standard addition of FF A to
soybean and corn oil and by analyzing AOCS Smalley check samples, which inc1uded
five types of oils (crude coconut, crude corn, crude safflower, cottonseed, and marine
oils). For the standard addition experiments, an FFA mixture was prepared by
saponifying olive oil with 50% w/v KOH followed by titration with 6M HCI to
regenerate the FF A (13), extraction into n-hexane, removal of the solvent using a rotary
evaporator, and titration of the residue to determine %FF A (expressed as % oleic acid).
Soybean and corn oils were then spiked (w/w) with six levels of this FF A mixture. These
samples and the Smalley check samples were analyzed for their FF A content by the FTIR
method as well as by the AOCS titrimetric method. Reproducibility was evaluated as the
standard deviation around the mean oftriplicate analyses (SDr). For the standard addition
experiments, accuracy was àssessed in terms of mean difference (MDa) and standard
deviation of the differences (SDDa) with respect to the gravimetrically spiked amounts of
FF A. In the case of the Smalley check samples, the FTIR results were compared to the
results of certified laboratories that had analyzed the samples by the AOCS titrimetric
procedure, using the statistical data (mean, min and max) provided with the samples.
3.4. RESULTS AND DISCUSSION
3.4.1. Analytical Concepts
As noted, FTIR methods developed to date for FF A analysis have been either
matrix dependent or l?rone to saponifica~ion errors. In FTIR methodology developed for
AN analysis in lubricants, the weak baSe K-phthal, a dicarbonyl compound, served as a
signal-transducing reagent by reacting with all organic and inorganic acids present to
form a single product, phthalimide, allowing AN to be determined from the intensity of
phthalimide's strong v (C=O) band at 1729 cm-l, with differential spectroscopy then
being used to eliminate the oil matrix effects (12), This method is not directly applicable
to edible oils, because the 1729 cm-l phthalimide band is masked by overwhelming
absorptions of the ester linkages of the TAG, whereas the second phthalimide v (C=O)
band at 1773 cm-l is too weak to measure FFA concentrations of <1 %. However, for the
analysis of FF A, signal transduction is not required per se, only the stoichiometric
conversion of FF A to their salts, which can then be quantitated directly by measurement
44
of their v (COO-) absorption. Moreover, the base used should be too weak to hydrolyze
TAG, unlike the KOH employed in previous work (6,10). K-phthal (PKa = 9.9) was found
to meet this criterion, with the additional advantage that I-PrOH could be used to deliver
this reagent into oils, thereby eliminating phase separation, with the FF A salts remaining
soluble in the I-PrOHioil mixture. The stoichiometric reaction of K-phthal with FF A is
presented below:
RCOO H :::::;;~r==:::!"'~
o
e® NH + RCOO K
o
As in the lubricant AN methodology, the problem of matrix effects, which may
arise in analyzing different oil types or result from the presence of various minor
constituents in the samples analyzed, is addressed by utilizing differential spectroscopy.
This approach involves splitting the sample into two parts; one part is then treated with
K-phthal in I-PrOH (designated RR) while the other portion is treated only with an
equivalent amount of I-PrOH (designated BR) and serves as a reference for the reacted
sample. Since the spectral features of the oil are invariant in the spectra of the BR- and
RR-treated portions, 'they are cancelled out in the differential spectrum and only the
spectral changes associated with the aeid/base reaction are left for evaluation (Figure
3.2).
3.4.2. Calibration and Stability of the Reaction
Figure 3.2 illustrates typical differential spectra obtained when calibration
standards (soybean oïl spiked with hexanoic aeid over a range of 0-4% FFA) were
analyzed using the protocol described above. As the amount ofhexanoic acid increases in
each subsequent standard, the v (COO) signal at 1570 cm-1 and that of the phthalimide v
(C=O) band at 1773 cm-1 rise concurrently. The loss of the v (C=O) band of hexanoic
acid, which would manifest itself as a strong negative band around 1710 cm-l, is largely
45
:/"'--', lost in the noise (1770-1700 cm- l) resulting from the subtraction of the off-scale bands of
the ester linkage of the oil.
1570 cm-1
1.0 1773 cm- l
0.8
~ y 0.6 = = ~ .. 0 Vi 0.4 ..c <
0.2
0.0~~r11
1800 1750 1700 1650 1600 1550
Wavenumber (cm-1)
/ i f
Figure 3.2. Potassium phthalimide as a reagent to carry out the acid/base reaction. Spectra were recorded in a 500 Ilm cell at 8 cm-l resolution. The bands at 1773 and 1570 cm- l are due to the phthalimide and hexanoate, respectively, formed in the reaction. The region between 1760 and 1710 cm- l is obscured by noise in the differential spectrum because of the intense oïl absorption in this region.
A mean plot with SD bars for three sets of hexanoic acid calibration standards
obtained by measuring the v (COO-) band at 1570 cm-l referenced to 2150 cm-1 vs. FFA
concentration expressed as % oleic acid is presented in Figure 3.3. The overalllinear
regression equation obtained for the composite calibration was:
46
% FF A(OIeic) = 4.281 * A(1570/2150) - 0.0436
R2 = 0.999 SD = 0.020 [3.1 ]
Figure 3.3. Composite calibration curve obtained from the differential spectra of three independent sets of hexanoic acid-spiked soybean oil standards of the type illustrated in Figure 3.1.
The linearity of the composite calibration plot, with an intercept well within 3 x
the regression SD, and the overall SD of -0.02% FF A indicate that the three individual
calibrations that were performed gave highly consistent results. However, it was noted
that calibrations tended to drift over a period of a week, and this was found to be due to
changes in the K-phthal reagent, making it necessary to perform a reagent blank
correction. Although K-phthal is practically insoluble in I-PrOH and is delivered as a
suspension, small amounts do solubilize slowly over time. Because the solubilized K
phthal has absorptions that overlap with the v (COO) band used to quantitate the FFA
salts, it can introduce a measurable bias into FF A measurements over time, albeit taking
several weeks to develop in a freshly prepared reagent. To account for this reagent
47
background signal, a mineraI oil, which does not contain any FF A, is run as a blank to
compensate for absorptions contributed by any solubilized K-phthal. The apparent FFA
contribution of the blank is subtracted from the values obtained for all oil samples
subsequently analyzed to account for any changes in K-phthal concentration.
Duplicate analyses of calibration standards were conducted 24 h apart to confirm
that K-phthal does not attack TAG. Based on the MDr and SDDr of 017% and 0.024%,
respectively, it was conc1uded that no significant saponification took place over that time
period. This finding was corroborated by similar studies with samples of various refined
oils.
3.4.3. Evaluation and Validation
To establish the efficacy of the methodology developed, standard addition
experiments were carried out by adding a pre-prepared mixture of FF A obtained from
olive oil to both refined corn and soybean oils, which were then analyzed in triplicate by
the FTIR and the titrimetric AOCS method. The data obtained using these two methods
are summarized in Table 3.1, with the accuracy ofboth methods being assessed by using
the gravimetrically added amounts ofFF A as the reference values.
In terms of overall reproducibility, the FTIR method had a mean SDr of 0.029%,
slightly better than that of the AOCS procedure (0.038%). In terms of overall accuracy
relative to gravimetric standard addition, both methods have small positive biases,
slightly greater than. the overall SDr, indicating that both the soybean and corn oils , " " ,
contained traces ofFFA prior to standard addition. For each method, the value of SDDa,
which is a measure of the variability around the MDa, is of the same order of magnitude
as the SDr value, with the FTIR method again performing slightly better. Figure 3.4
present a composite plot of all the FF A results obtained for both oils by both methods
against the reference gravimetric data. The regression equation for the composite data
illustrated in Figure 3.4 is:
FFA(IRITitration) = 0.99996 FFA(StandardAddition) + 0.047
R2 = 0.999 SD =0.030 [3.2]
48
Table 3.1. Results of Triplicate Analyses ofOils Spiked with Various Amounts an FF A Mixture by the AOCS Titrimetric and FTIR Methodsa
FFA spiked AOCSMethod FTIRMethod Oil
(%W/W)b Mean SD Mean SD
0.000 0.028 0.003 0.015 0.032
0.501 0.524 0.015 0.526 0.019
Soybean 1.002 1.063 0.034 1.042 0.035
1.508 1.543 0.061 1.549 0.031
2.003 2.033 0.029 2.056 0.032
2.490 2.531 0.055 2.505 0.037
0.000 0.057 0.011 0.066 0.020
0.500 0.546 0.009 0.573 0.027
Corn 0.999 1.120 0.053 1.038 0.023
1.475 1.573 0.068 1.512 0.005
2.002 2.094 0.076 1.991 0.088
2.502 2.578 0.051 2.548 0.002
Mean SDr 0.038 0.029
MDa 0.059 0.036
SDDa 0.030 0.022
"Abbreviations: SDr. SD for reproducibility; MD •• mean difference for accuracy; SDD •• SD of the differences for accuracy.
bExpressed as % oleic acid.
The plot and regression equation illustrate the excellent concurrence between the
titrimetric and FTIR methods as well as their comparable ability to track the amounts of
FF A added gravimetrically to the two oils.
49
3.0 • AOCS Soybean oil • AOCS Comoil -.. 2.5 ! 6. FTIR Soybean oil
~ * FTIRComoil
~ 2.0 = --"CI ~ Jo.. 1.5 = c:I.l ~ ~
~ 1.0 -< ~ ~ 0.5
0.0 • 0.0 0.5 1.0 1.5 2.0 2.5
FFA Added (% w/w)
Figure 3.4. Graphical comparison of results obtained by the AOCS and FTIR methods for the oils that were subjected to standard addition of FF A mixture, relative to the gravimetrical data.
To further validate the performance of the FTIR method, five AOCS Smalley
check samples were analyzed in triplicate. The results obtained relative to the analytical
data provided with these oïl samples are presented in Table 3.2. In general, there is , ,
excellent concurrence between the FTIR' mean and Smalley mean FF A values except for
crude coconut oil. Analysis of the Smalley samples in our laboratory by the AOCS
method were also in line with the Smalley means, except again for crude coconut oil,
which produced a value of 0.312 ± 0.005, very much in line with the FTIR result
obtained, suggesting that the FF A content of this sample had changed in the time that had
elapsed since its analysis in laboratories participating in the Smalley check sample
program. Considering the results for the Smalley check samples as weIl as those obtained
by standard addition, it is evident that the FTIR method is capable of producing accurate
and reproducible FF A data independent of oil type and appears to be a valid alternative to
the AOCS titrimetric procedure.
50
Table 3.2. Results of Triplicate Analysis ofSmalley Check Samples by AOCS Method and FTIR Methoda
AOCS Method FTIRMethod Oil sample
Max Min Mean SD nb Mean SD
Crude safflower oïl 0.500 0.330 0.457 0.036 18 0.467 0.057
Cottonseed oil 0.180 0.040 0.130 0.053 8 0.156 0.058
Crude coconut oil 0.200 0.080 0.136 0.029 20 0.297 0.014
Crude corn oïl 1.720 1.120 1.419 0.124 19 1.431 0.019
Marine oïl 1.970 1.800 1.864 0.044 12 1.810 0.024
aResults expressed as % oleic acid (w/w)
b n = number of laboratories that reported results for each sample.
The FTIR methodology for FF A analysis developed in this work combines the
respective advantages of the direct and indirect approaches previously described in the
literature and overcomes their limitations. As in other indirect approaches, enhanced
sensitivity by comparison with direct measurement of the FF A v (C=O) band is achieved
by reacting the FFA with a base and measuring the v (COO") band of the salt formed.
However, a key difference is the use of a K-phthal suspension in 1-PrOH as a reagent
instead of the KOHImethanol reagent previously used (6,10) as its weakly basic . ,
properties provide a means of avoiding sàponification. Furthermore, the extraction step
previously required in the indirect FF A methods is eliminated because no phase
separation occurs and the FF A salts remain soluble in the I-PrOHIoil mixture. The use of
differential spectroscopy provides a means of minimizing matrix effects by canceling out
the spectral contributions of the oil and contaminants therein, and spectral interferences
arising from the slight solubilization of K-phthal in a non-freshly prepared reagent are
compensated for by performing a reagent blank correction. These combined procedural
elements allow a calibration to be developed by utilizing gravimetrically prepared
standards of hexanoic acid in oil. The net result is that the FTIR method developed is a
primary method, independent of other methods for calibration. It is also rapid and simple
51
to execute, and the FTIR portion of the analysis has been automated by integrating the
spectrometer with an autosampler (Thermal-Lube Inc., Pointe-Claire, QC, Canada),
allowing for the analysis of up to 60 paired samples per hour. Given that only
measurements at two wavelengths are needed for this analytical procedure, a simple dual
wavelength filter-based IR instrument could also be used. With these hallmarks of a
sound general method, this new methodology could serve as the basis for an AOCS
instrumental method for the analysis ofFFA in edible oils.
3.5. ACKNOWLEDGMENT
The authors would like to thank Sultan Qaboos University, Oman, for support of
Mr. Alawi and the Natural Sciences and Engineering Research Council (NSERC) of
Canada for financial support of this research. The technical assistance of Ms. Claudia
Mucciardi is also acknowledged.
3.6.
1.
2.
3.
4.
5.
6.
7.
8.
REFERENCES
van de Voort, F. R, Sedman J. and Russin, T. Lipid Analysis by Vibrational Spectroscopy, Eur. J. Lipid Sci. Technol. 103,815-826 (2001).
O'Brien, R D. Fats and Oils: Formulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998.
Official Methods and Recommended Practices of the American Oil Chemists' Society, 4th edn., 1989, Champaign, AOCS Official Method Ca-5a-40.
Determination of the Free Fatty Acids in European Commission Regulation (EED) No. 2568/91, Official Journal of the European Communities, No. L 248 (5.9.91), p. 6. '
, ' "
Lanser, A C., List, G. R, Holloway, R K., and Mounts, T. L. FTIR Estimation of Free Fatty Acid Content in Crude Oils Extracted from Damaged Soybeans, J Am. Oil Chem. Soc., 68, 448-449 (1991).
Ismail, A. A., van de Voort, F. R and Sedman, J. Rapid Quantitative Determination of Free Fatty Acids in Fats and Oils by FTIR Spectroscopy. J Am. Oil Chem. Soc., 70, 335-341 (1993).
Vedeyen, T., Verhe, R, Cano, A, Huyghebaert, A and De Greyt, W. Influence of Triacylglycerol Characteristics on the Determination of Free Fatty Acids in Vegetable Oils by Fourier Transform Infrared Spectroscopy, J Am. Oil Chem. Soc., 78, 981-984 (2001).
Man, Y. B., Moh, M. H. and van de Voort, F. R Determination of Free Fatty Acids in Crude Palm Oil and Refined-Bleached-Deodorized Palm Olein Using
52
. ~.
Fourier Transfonn Infrared Spectroscopy, J. Am. ail Chem. Soc., 76, 485-490 (1999).
9. Bertran, E., Blanco, M., Coello, J., Iturriaga, H., Maspoch, S. and Montoliu, 1. Detennination of Olive Oil Free Fatty Acid by Fourier Transfonn Infrared Spectroscopy, J. Am. ail Chem. Soc., 76, 611-616 (1999).
10. Cafiada, M. J. A., Medina, A. R and Lendl, B. Detennination of Free Fatty Acids in Edible Oils by Continuous-Flow Analysis with FT-IR Spectroscopic Detection, App!. Spectrosc., 55, 356-360 (2001).
11. van de Voort, F. R, Sedman, J., Yaylayan, V. and Saint Laurent, C. The Detennination of Acid Number and Base Number in Lubricants by FTIR Spectroscopy. Appl. Spectrosc., 57, 1425-1431 (2003).
12. Kishore N. R A., Guide to ASTM Test Methods for the Analysis of Petroleum Products and Lubricants (ASTM, West Conshohocken, PA, 2000), p. 45~
13. Official Methods of Analysis of the Association of Official Analytical Chemists, 17th edn., Arlington, 2000, Method 972.28 .
53
...--., l '
BRIDGE
In Chapter 3, an FTIR method for determination ofFFA content in edible oils was
developed by solubilizing oils in I-PrOH containing a weak base and spectrally
evaluating the FF A salts formed, using differential spectroscopy to minimize matrix
effects. The method focused on finding ways to avoid or eliminate problems that were not
overcome by previously developed FTIR methods, such as saponification and calibration
dependency on oil type. Although successful in achieving these goals and providing a
reliable and reproducible instrumental method for the determination of FF A, this new
FTIR method is not significantly better than standard titrimetric procedures in terms of
sensitivity. In Chapter 4, the knowledge and experience gained from this initial work is
used to develop a much more sensitive method which allows for the analysis of low
levels ofFFA in refined edible oils .
54
CHAPTER4
A NEW METHOD FOR THE ANALYSIS OF LOW LEVELS OF FFA IN
REFINED EDIBLE OILS
4.1. ABSTRACT
This paper summarizes the application of stoichiometric analytical approaches to
quantitative IR analysis and describes the development of a rapid and sensitive Fourier
transform infrared (FTIR) method using such an approach for the determination of low
levels «0.005%) of free fatty acids (FFA) in refined edible oils. The method simply
involves mixing the sample with methanol containing 2 gIL sodium hydrogen cyanamide
(NaHNCN) on a vortex mixer for 30 s to convert the FFA to their salts, centrifuging the
sample to separate the methanol phase containing the FF A salts from the oil, recording
the FTIR spectrum of the upper methanollayer in a 100 Jlm CaF2 transmission flow cell,
and ratioing this spectrum against that of the NaHNCN/methanol solution. The
concentration of FF A salts is determined from the resulting differential spectrum by
measurement of the v (COO) absorbance at 1573 cm-l relative to a reference wavelength
of 1820 cm-l. A calibration spanning the range 0-0.1 % FF A (expressed as oleic acid) was
devised by gravimetric addition of a defined, pure fatty acid to an acid-free oil.
Validation of the method by standard addition of palmitic acid to a variety of oils yielded
an overall standard error of <±0.001% FFA. Comparison oftriplicate FTIR and IUPAC
titrimetric analyses of oils spiked with palmitic acid demonstrated that this FTIR method " , . ' '
was more sensitive, accurate and reproducible than the titration procedure, the latter
having a significant positive bias of ~0.02%. Solventloil consumption in the FTIR
method is 2 mLll 0 g vs. 150 mL/20 g for the titrimetric procedure. The FTIR method
developed is particularly well suited for the determination of the low levels of FF A in
refined oils but can readily be adapted with a simple adjustment of the oi1:methanol ratio
to coyer FFA levels ofup to 4.0%.
4.2. INTRODUCTION
The application of chemometric techniques has represented a major advance in
quantitative IR analysis. In particular, partial-Ieast-squares regression (PLS) has been of
55
widespread utility (1) as it provides a powerful means of extracting quantitative spectral
information related to component(s) of interest from the spectra of complex samples by
mathematically modeling partially overlapping bands or other sources of spectral
interference. However, as the complexity of the system increases, so does the difficulty of
developing reliable PLS calibration models and we have encountered diverse
circumstances where PLS proved to be unsatisfactory as a calibration technique.
Consequently, an alternative approach to quantitative IR analysis was developed based on
the use of a reagent that reacts stoichiometrically with the component(s) of interest to
pro duce a readily measurable IR signal. Furthermore, by combining this "signal
transduction" approach with differential spectroscopy, quantitation can often be achieved
via a simple univariate calibration.
Table 4.1 illustrates a number of successful applications of this approach. In the
first case, difficulties were encountered in developing a robust PLS calibration model for
determination of the peroxide value (PV) of edible oils. PV is a measure of
hydroperoxides (ROOH) , the primary oxidation products in edible oils, but FTIR
quantitation of these species was complicated by both spectral overlap and hydrogen
bonding interactions with a large variety of secondary oxidation products and other
components that may be present in edible oils (2). These problems were overcome by
taking advantage of the rapid stoichiometric reaction of hydroperoxides with
triphenylphosphine (TPP) to form triphenylphosphine oxide (TPPO), which has an
isolated and intense absorption band ~at al~owed for the accurate determination of PV " ,
down to <1.0 mequivalents ROOHlkg oil (3).
The next three examples in Table 4.1 concern the analysis of new and used
lubricating oils, which is beset by even more complex matrix effects than those
encountered in oxidized edible oils. In the case of moi sture analysis (4), signal
transduction was carried out by stoichiometrically converting dimethoxypropane through
its reaction with water into acetone, a strong IR absorber, providing an alternative to the
problematic Karl Fischer titration procedure. For the determination of acid number (AN)
and base number (BN) in lubricants, there was the additional complication of the
"structural specificity" of IR analysis as opposed to the "chemical specificity" of the
traditional titrimetric methods employed for these analyses. The latter methods directly
56
VI -....l
')
Table 4.1. FTIR Analyses Based on the Use ofStoichiometric Reactions for "Signal Transduction"
Analysis· Reagent Stoichiometric Reaction FTIR Measurement
PV (>1.0) Triphenylphosphine ROOH + TPP ~ ROH + TPPO 542 cm-1 [X-sensitive ring-breathing
Edible Oils (TPP) in l-hexanol vibration ofphenyl groups in TPPO]
H20 «1000 ppm) 1,3-Dimethoxypropane 2H20 + DMP ~ CH3COCH3 + 2CH30H + H20 1717 cm-1 [v(C=O) absorption of
Lubricants (DMP) in isooctane - acetone]
AN «0.1 mg KOHlg) Potassium phthalimidè - HA + K+Phthar ~ Ph + A" K+ 1774 or 1727 cm-1 [v(C=O)
Lubricants (K-Phthal) in 1-PrOH' - absorptions of phthalimide]
-
BN «0.1 mg KOHlg) Trifluoroacaetic acid B: + TFA ~ TFA"BH + 1679 cm-1 [v(COO") absorption of
Lubricants (TF A) in 1-PrOH trifluoroacetate]
FFA (>0.2% C 18:1) Potassium phthalimide RCOOH + K+Phthar ~ RCOO- + Ph 1570 cm-1 [v(COO") absorption of
Edible Oils (K-Phthal) in 1-PrOH RCOO-] or 1776 cm-1 [v(C=O)
absorption of phthalimide]
FFA(>0.002% C 18:1) Sodium hydrogen RCOOH + NaHNCN ~ RCOO- + H2NCN 1573 cm} in spectrum ofmethanol
Refined Edible Oils cyanamide in methanol extract [v(COO") absorption of
RCOO1 - --~~
apV = Peroxide Value, H20 = Moisture, AN = Acid Number, BN = Base Number, FFA = Free Fatty Acids.
Rer.
(2)
(3)
(4)
1
(5)
(6)
This work
measure a large variety of acids and bases, their response being dependent only on the
pKa of the acid or base in relation to the titrimetric endpoint. In contrast, prediction of AN
or BN by direct IR analysis would require the development of calibrations that model all
the species contributing to the acidiclbasic characteristics of lubricating oils, many of
which are undefined. This severe limitation was overcome by reacting all the acidic or all
the basic species with, respectively, a basic or an acidic "signal-transducing" reagent. By
subtracting the spectrum of the unreacted sample from that of the reacted sample, the
spectral changes associated with the acid-base reaction were isolated, and the extent of
conversion of the "signal-transducing" reagent could be directly measured to determine
the AN or BN of the sample (5).
The last two examples cited in Table 4.1 concern the application of similar
concepts to the determination of the free fatty acid (FF A) content of edible oils, the first
of which was described in a previous paper (6). Among the oil quality parameters, FF A
content is a crucial factor associated with the quality and economic value of edible oils,
especially for unrefined high value oils such as olive oil. FF A content is also an important
quality indicator in relation to oil processing and is used to assess deodorizer efficiency
or as an indicator of frying oil quality (7,8). FFA content is most commonly determined
by titration of an oil, dissolved in neutralized ethanol or ethanol/diethyl ether, with a
strong base to a phenolphthalein endpoint (9,10). Although the standardized titrimetric
methods are fairly sensitive, with limits of detection (expressed as percent oleic acid) on
the order of 0.03% being attainable, ,mor~.sensitive methods would be useful for the \' "i f
analysis of refined, bleached and deodorized (RBD) oils, which tend to have FF A levels
of :::;0.05% (7), and could also provide an alternative means of monitoring secondary
oxidation products (11) accumulating in an oil in the form of carboxylic acids. In recent
years, a variety of approaches have been investigated as possible alternatives to the
titrimetric methods employed to determine the FF A content of oils, inc1uding the use of
flow injection systems (12,13) pH metric, potentiometric and colorimetric (14-16)
methods, and chromatographic procedures (17-20) as well as FTIR-based spectroscopic
techniques (21-28). Although many of these offer substantial benefits, in terms of speed
of analysis, amenability to automation, andlor a reduction in the use of solvents and the
attendant environmental problems and disposaI costs, none of them provide a substantive
58
~'" gain in sensitivity over that attained with titrimetric methods. This paper describes a
simple, robust FTIR method based on the concepts outlined above that is capable of
measuring FF A levels as low as 0.005% in refined oils. Thus, the new methodology
described in this paper exemplifies the means by which the sensitivity of IR analysis can
be substantially enhanced under certain circumstances by the signal
transductionldifferential spectroscopy approach.
4.3. METHODOLOGY
Reagents and Standard Methods. Sodium hydrogen cyanamide (NaHNCN,
99+%), palmitic acid (99%), and anhydrous methanol (MeOH) were obtained from
Aldrich (St. Louis, MO) and were all of analytical grade. Refined edible oils were
purchased locally or obtainedfrom Canamera Foods (Toronto, ON, Canada). The reagent
solution employed in the FTIR FF A analysis was prepared by dissolving NaHNCN in
anhydrous MeOH (2 gIL). This solution was allowed to stand for -4 days or until the v
(C=N) band at 2100 cm-1 completely disappeared before use.
Instrumentation. The FTIR spectrometer used for this study was a Bornem
WorkIR (Bornem, Quebec, PQ, Canada) equipped with a DTGS detector and purged with
dry air using a Balston dryer (Balston, Lexington, MA). The sample-handling accessory
was a valved 100 p,m CaF2 transmission flow cell (Dwight Analytical, Toronto, ON,
Canada). Samples were aspirated into the cell under vacuum, and the cell was flushed
c1ean after each sample with 1 mL of methanol. All spectra were collected by co-adding
32 scans at a resoluti~n of 8"cm- l and' a ~g~in of 1.0. The spectrometer was controlled by
an IBM-compatible Pentium ISO-MHz PC running under proprietary Windows-based
UMPIRE® (Univers al Method Platform for InfraRed Evaluation) software (Thermal
Lube, Pointe-Claire, PQ, Canada). This software provides programming capabilities so
that repetitive operations can be performed in a specified sequence and designated
spectral data collected and processed through a calibration equation, thereby automating
the analysis to provide direct output ofFFA data.
Preparation of Calibration Standards. A series of 12 standards covering a range
of 0-0.1 % FF A was prepared by gravimetric addition of palmitic acid to a refined and
deodorized soybean oil after it had been ron though an activated silica gel column to
59
remove any traces of FF A and other oxygenated compounds. The FF A contents of the
standards were expressed in terms of % oleic acid.
Sample Preparation. Ten grams (± 0.001 g) of each oil sample or standard was
weighed on an analytical balance into a tared 15-mL clinical centrifuge tube. To the tube
containing the oil, 2 mL of the NaHNCN reagent solution was added using a calibrated
re-pipette. The tubes were capped, shaken on a vortex mixer for 30 s, and then
centrifuged for 5 min at 6000 rpm (-5000xg) to ensure consistent separation between the
oil and methanol phases.
Analytical Protocol. Approximately 1 ml of the methanol reagent was loaded into
the transmission flow cell and its single-beam spectrum recorded to serve as the
background spectrum. A new background spectrum was collected in the same manner
after every 20 samples or 1 h, whichever occurred tirst. For both samples and calibration
standards, 1 mL of the upper methanollayer formed after centrifugation was loaded into
the cell and its single-beam spectrum was recorded and ratioed against the
NaHNCNlMeOH background spectrum. The peak height of the carboxylate band at 1573
cm- l was then measured relative to an invariant baseline point at 1820 cm- l. The overall
sample preparation procedure and analytical protocol is illustrated in Figure 4.1.
Calibration and Validation. The calibration standards were taken through the
analytical protocol described above, and a calibration equation for the prediction of FF A
content was derived by plotting the concentrations of the standards (% oleic acid) vs. " , .' l 'I , ,
carboxylate peak height. The reproduCibÜity and accuracy of the FTIR method were
assessed by standard addition, spiking three different acid-free oils (canola, soybean and
sunflower) with known amounts ofpalmitic acid (w/w). These oils were each analyzed in
triplicate, on different days, by both the IUP AC titrimetric method (9) and the FTIR
method to allow a direct comparison of their performance. A comparative analysis of
locally purchased retined oils (soybean, sunflower, peanut, and corn oils and a
commercial oil blend) was also carried out.
4.4. RESULTS AND DISCUSSION
/""', The various approaches that have been investigated for the determination of FF A
content in edible oils by FTIR spectroscopy were reviewed in a previous paper (6), and
60
-~---/
~ 2ml -MeOHlNaHNCN
~ 10g
1 ( Mix )
~
( Centrifuge ) FTIR 1 -(~) (Background) .........
l Spectrum 2 - Spectrum 1 (--A-) ./
FTIR 2 -(~) - (Sample)
Figure 4.1. Schematic diagram of the sample preparation and analytical protocol.
the limitations of each approach discussed. In that paper, these limitations were addressed
by the development of a new FTIR method based on previous work on AN determination
in lubricants, which employed the mild base potassium phthalimide as a signal
transducing reagent i~ conjunction wiili differential spectroscopy to circumvent matrix
effects (6). Although the phthalimide FTIR procedure is bOth more accurate and
reproducible than conventional titration, it is ultimately not more sensitive per se. This is
largely due to the use of propanol as a solvent and polarity enhancer for the reaction,
effectively diluting the COO- IR signal. A means by which sensitivity could be improved
would be to treat the oil with methanol containing a base which is immiscible with the oïl
to facilitate the acid-base reaction as well as concentrate the FF A salts in the methanol
layer (14). Such a procedure would have the additional advantage of minimizing matrix
effects by partitioning out the spectral contribution of the oil. For highly accurate
analyses, a weak base would be required to avoid any saponification of the oil and one
61
could use either the spectral changes associated with the loss of the base or formation of
FF A salts as a basis for quantitation. Examination of a range of reagents led to the
consideration ofthe sodium salt of carbodiimide (NaHNCN), which has a strong v (C=:N)
absorption at 2100 cm-l, is readily soluble in methanol, and is capable of converting
FF As to their carboxylate saIts but not capable of saponifying triacylglycerols. In the first
instance, the proportionate decrease in the intensity of the v (C=:N) band appeared to be a
very good measure of the amount of FF As spiked into oils. Although this reaction
worked, its was found that the reagent spectrum was unstable and that the band at 2100
cm-l slowly disappeared over a period of - four days with the concomitant appearance of
two new bands at 1650 and 1610 cm-l (Figure 4.2). Based on assignment of these two
bands to C=N stretching and NH bending vibrations, respectively, the structural
rearrangement from NaHN-C=:N -7 NaN=C=NH was postulated to be taking place in
solution over time. This conversion was conftrmed spectrally by dissolving NaHNCN in
methyl alcohol-d (MeOD), whereby the band at 1610 cm-l
shifted about 100 cm-l to
lower frequency owing to hydrogen-deuterium exchange.
0.30 B----"-1\
0.15 A
~ ~ = = ..c 0.00 ... 0 f'-l
..c <
-0.15
B
-0.30 2400 2200 2000 1800 1600
Wavenumber (cm-1)
Figure 4.2. Differentiai spectra for sodium hydrogen cyanamide in methanol. The spectra show the decomposition of the reagent over time. A = one day; B = four days.
62
This molecular rearrangement, which was established to be complete within 4
days after preparation of the MeOH/NaHNCN reagent, did not affect its ability to convert
FF As to their respective salts without causing oil saponification. The reactivity of the
converted MeOHINaHNCN solution as weIl as its spectral characteristics remained
stable, with solutions kept at room temperature being used for up to two months without
any apparent deterioration in their efficacy. On the other hand, the measurement of the
. C=N band originaIly envisioned as a basis for quantitation is lost as a result of this
transformation. However, measurements made using the carboxylate band ofFFA salts at
1573 cm-I were both reproducible and responsive to low FFA levels because of the
concentration of the FF A salts in the methanollayer and the high extinction coefficient of
the carboxylate band. It was found that optimum sensitivity and reproducibility were
achieved with a 5:1 oil:methanol ratio (10 g oil plus 2 mL of "aged" MeOH/NaHNCN
reagent mixed in a standard 15-mL clinical centrifuge tube), producing consistent
reactions and reproducible separations of the MeOH and oil phases when centrifuged for
5 min at 5000 x g. Oil stability with respect to saponification by the reagent was assessed
by incubating oil samples with the reagent for 24 h and did not result in any measurable
oil hydrolysis, but did lead to a minor displacement effect (~0.01% FFA) due to sorne
additional oil migration into the methanollayer over extended periods oftime.
4.4.1. Calibration
A calibration curve was developed by using a set of standards covering the range
of 0.0 to 0.1 % FFA, prepared by adding palmitic acid to acid-free soybean oil. Figure 4.3
illustrates the differential spectra obtained for these standards using the optimized
analytical protocol outlined in Figure 4 1. The carboxylate anion produced and extracted
into the Me OH layer shows a rising absorbance at 1573 cm-l, the other spectral features
at 1650 cm- I and the split band covering 1760-1700 cm- I representing the v (C=N)
absorption of the base and the v (C=O) ester linkage absorption of a smaIl amount of
solubilized oil. A plot of the absorbance measured at 1573 cm -l, referenced to a single
point baseline at 1820 cm-l, vs. % FF A (oleic acid) is presented in Figure 4.4. Linear
regression of the data obtained resulted in the following relationship:
63
1573 cm-t
0.16
~ 0.12 CJ ::1
= .c .. 0 f'-.l
0.08 .c -<
1820 cm-1
0.04
1800 1750 1700 1650 1600 1550 1500
Wavenumber (cm-1)
Figure 4.3. DifferentiaI spectra for soybean oil spiked with palmitic acid (0.0 - 0.1%1 after carrying out the acid/base reaction. Spectra were recorded in a 100 J.lm cell at 8 cmresolution.
, 'l,
, '
% FF A = 0.70292 A(1573/1820) - 0.00883
R2 = 0.9998; SD = 6.706xl0-4 [4.1]
The regression SD implies that FFA levels in the order of 1/1000th of a percent may be
measurable by this technique.
4.4.2. Validation
Validation and comparison of the FTIR method relative to the IUP AC titrimetric
method were carried out by analyzing three acid-free oils (soybean, canola and
sunflower) spiked with palmitic acid. Figure 4.5 presents comparative plots for triplicate
64
0.16
,.... e 0.12 u
co N 00 ... ~ 1/') • e, ~ 0.08 CJ
== • ~ .c ,.. 0 fil .c 0.04 • -< t • t
•• 0.00
0.00 0.02 0.04 0.06 0.08 0.10
0/0 FFA w/w (Oleic Acid)
Figure 4.4. Calibration curve for %FF A in oïl obtained from the differential spectra in Figure 4.2. The %FF A is expressed as % oleic aeid (w/w). Error bar amplitude indicates mean ± SD of three replicates.
analyses of these oïls by FTIR spectroscopy and the IUP AC titration procedure,
respectively. The corresponding regression equations are:
, ' 1,/
R2 = 0.9998 [4.2]
IUPACFFA = 1.047 FFA + 2.3xlO-2
R2 = 0.9887 [4.3]
These results clearly indicate that the FTIR method tracks standard addition very
weIl, with a slope and an intercept very close to unity and zero, respectively, with an
overall error of about <0.001 %. The titrimetric plot clearly shows greater variability.
65
0.12 -~ ~ 0.09 ~ C '-' ~ joooj
~ 0.06 ~ ~ .c < 0.03 ~ ~
0.00
0.12 -~ ~ ~ 0.09
= o .,.. ..... ~ !:: 0.06 .,.. ~ ~ .c < 0.03
~ 0.00
A
* I
• f • • .,;.
.. • •
0.00 0.02 0.04 0.06 0.08 0.10
FFA Added (%W/W)
B
0.00 0.02 0.04 0.06 0.08 0.10
FFA added (o/OW/w)
Figure 4.5. A plot of %FF A determined by the FTIR method (A) and %FF A determined by the titrimetric method (B) against the spiked amount. Error bar amplitude indicates mean ± SD of three replicates.
66
.~ ..
Table 4.2 presents the data in terms of the relative mean difference (MD) and
standard deviation of the differences (SDD) for both accuracy (a) and reproducibility (r).
The MDa of 0.025% for the titrimetric method indicates a significant bias relative to
standard addition considering its SDDa, while the FTIR results are basically an order of
magnitude better in terms of both accuracy and reproducibility with no significant bias.
The apparent titrimetric over determination relative to the FTIR method may be due to
systematic errors associated with a slower acid/base reaction owing to the lower solvent
polarity, the substantive volume of solvent used, and C02 absorption, as well as
variability contributed by the visual endpoint determination. Table 4.3 presents
comparative triplicate FTIR and titrimetric data obtained for locally purchased samples of
refined oils. The two methods correlate (R2 = 0.83), showing a similar trend as the
standard addition data, with the titrimetric procedure predicting higher values
individually as well as an overall mean bias of ....Q.02% and yielding poorer
reproducibility.
Table 4.2. Mean Difference (MD) and Standard Deviation of the Differences (SDD) for Accuracy (a) for the Titrimetric and FTIR Procedures vs. Gravimetrie Addition and for Reproducibility (r) for Duplicate Analyses Carried Out by Titrimetric and FTIR Analyses, Respectively
Statistic FfIR Titration
MDa 5.84 X 10-4 2.50 X 10-2
" : 1',111 ,! f
SDDa 7.00 x 10-4 5.10 X 10-3
MDr 4.56 x 10-4 3.65 xlO-3
SDDr 9.47 x 10-4 9.86 X 10-3
The carbodiimide FTIR method was further examined in relation to expanding its
general utility by changing its oil:reagent ratio. By decreasing the oil:reagent ratio from
5:1 to 1:4, while maintaining the same carbodiimide concentration, a calibration covering
a range of 0-4% FFA was developed which produced the following calibration linear
regression equation:
67
% FFA = 18.994*A(1573/1820) - 0.03598
R2 = 0.9999 SD = 0.0206 [4.4]
Thus simply by changing oil:reagent ratio, semi-refined and crude oils can also be
analyzed accurately using the same basic analytical protocol outlined in Figure 4.1. As
such, the method developed provides a simple FTIR-based analytical procedure for the
measurement of both very low and moderate levels of fatty acids in oils. Aside from
providing excellent sensitivity and reproducibility, it overcomes many of the limitations
associated with conventional titrimetric and potentiometric methods. There is no need to
use an FTIR spectrometer for this analysis per se, given that measurements at only two
wavelengths are required and a simple dual-wavelength fiUer instrument would suffice.
ln either case, the method is readily amenable to automation.
Table 4.3. Results of Triplicate Analyses of Locally Purchased Oil Samples by the IUP AC Reference (Titrimetric) Method and the FTIR Method
Titrimetric method FTIRmethod Oil type
%FFA(w/wt SD %FFA(w/w)a SD
Peanut oil 0.050 5.39xlO-3 0.030 7.03xl0-5
Sunflower oil 0.041 ., .," 6.40xl0-3 0.021 9.43xlO-4
Comoil 0.053 3.50xl0-3 0.024 4.39xlO-4
Mixed oil 0.061 7.15xl0-3 0.032 8.28xl0-4
Soybean oil 0.045 3.62xlO-3 0.020 1.62xlO-4
Overall Mean 0.050 0.026
a Expressed as % oleic acid
68
I~'
4.5. CONCLUSION
Our work on FTIR analysis of lubricating oils led us to develop a novel approach
for the determination of acidity in non-aqueous systems based on the combined use of
signal transduction via a stoichiometric reaction and differential spectroscopy. In
subsequent work, we demonstrated the suitability of this approach for the quantitation of
FF As at levels of >0.2% oleic acid by employing the same reagent as used for the
determination of acidity in lubricating oils. In the present study, the use of sodium
carbodiimide and a modified procedure has allowed the analysis to be extended down to
FF A levels as low as 0.001 %, which would not be measurable by direct measurement of
FF A absorptions in the IR spectra of oils. With a simple adjustment in the reagent:oil
ratio the method becomes scalable and the analysis of semi-refined and crude oils
containing up to 4.0% FF A is possible. The strength of the method is its sensitivity and
allows for the possibility of using FF As as an indicator of lipid oxidation and to more
accurately monitor the refining processes carried out on edible oils.
4.6. ACKNOWLEDGMENT
F. R van de Voort acknowledges the financial support of the Natural Sciences
and Engineering Research Council of Canada and the cooperation of Canamera Foods for
providing oil samples for analysis. A AI-Alawi thanks Sultan Qaboos University for
financial support for his PhD studies.
4.7. REFERENCES , "
J" t
1. Haaland, D. M. Quantitative Infrared Analysis of Borosilicate Films usmg Multivariate Statistical Methods. Anal. Chem., 60, 1208-1217 (1988).
2. van de Voort, F. R, Ismail, A. A, Sedman, J., Dubois, J and Nicodemo, T. The Determination of Peroxide Value by Fourier Transform Infrared (FTIR) Spectroscopy. J. Am. Oil Chem. Soc., 71, 921-926 (1994).
3. Ma, K., Van De Voort, F. R, Sedman, J. and Ismail, A. A Stoichiometric Determination of Hydroperoxides in Fats and Oils by Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 74, 897-906 (1997).
4. van de Voort, F. R., Sedman, J., Yaylayan, V. and Saint Laurent, C. Determination of Acid Number and Base Number in Lubricants by Fourier Transform Infrared Spectroscopy. Appl. Spectrosc., 57, 1425-1431 (2003).
69
5.
6.
7.
8.
9.
10.
Il.
12.
13.
14.
15.
16.
17.
18.
19.
van de Voort, F. R, Sedman, J., Yaylayan, V., Saint Laurent, C. and Mucciardi, C. Quantitative Determination of Moisture in Lubricants by Fourier Transform Infrared Spectroscopy. Appl. Spectrosc., 58, 193-198 (2004).
AI-Alawi, A, van de Voort, F. R and Sedman, J. New Method for the Quantitative Determination of Free Fatty Acids in Oil by FTIR Spectroscopy. J Am. ail Chem. Soc., 81,441-446 (2004).
O'Brien, R D. Fats and ails: Formulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998
Senorans, F. J. and Ibanez; E. Analysis of Fatty Acids in Foods by Supercritical Fluid Chromatography. Anal. Chim. Acta, 465,131-144 (2002).
International Union Pure and Applied Chemistry. Standard Methods for the Analysis if ails, Fats and Derivatives, th revised and enlarged edn., Paquot, c., and Hautfenne, A., Eds. Blackwell Scientific, Oxford, 1987
Official Methods and Recommended Practices of the American ail Chemists' Society, 4th edn., 1989,Champaign, IL. AOCS Official Method Ca-5a-40
Velasco, J., Andersen, M. L. and Skibsted, L. H. Evaluation of Oxidative Stability of Vegetable Oils by Monitoring the Tendency to Radical Formation. A Comparison of Electron Spin Resonance Spectroscopy with the Rancimat Method and DifferentiaI Scanning Calorimetry, Food Chem., 85, 623-632 (2004).
Nouros, P. G., Georgiou, C. A. and Polissiou, M. G. Automated Flow Injection Spectrophotometric Non-Aqueous Titrimetric Determination of the Free Fatty Acid Content of Olive Oil. Anal. Chim. Acta, 351(1-3),291-297 (1997).
Mariotti, E. and Mascini, M. Determination of Extra Virgin Olive Oil Acidity by FIA-Titration. Food Chem., 73, 235-238 (2001).
Kwon, D. Y. and Rhee, J. S. A Simple and Rapid Colorimetric Method for Determination of Free Fatty Acids for Lipase Assay. J Am. ail Chem. Soc., 63, 186-189 (198(». , " ,
Tur'yan, Va. 1., Bel'ezin, O. Yu; Kuselman, 1. and Shenhar, A. pH-Metric Determination of Acid Values in Vegetable Oils without Titration. J Am. ail Chem. Soc., 73, 295-301 (1996).
Takamura, K., Fuse, T., Arai, K. and Kusu, F. A Review of a New Voltammetric Method for Determining Acids. J Electroanal. Chem., 468, 53-63 (1999).
Dermaux, A., Sandra, P. and Ferraz, V. Analysis of Free Fatty Acids and Fatty Acid Phenacyl Esters in Vegetable Oils and Margarine by Capillary ElectroChromatography. Electrophoresis, 20, 74-79 (1999).
Kotani, A., Kusu, F. and Takamura, K. New Electrochemical Detection Method in High-Performance Liquid Chromatography for Determining Free Fatty Acids. Anal. Chim. Acta, 465, 199-206 (2002).
Rosenfeld, J. M. Application of Analytical Derivatizations to the Quantitative and Qualitative Determination of Fatty Acids. Anal. Chim. Acta, 465, 93-100 (2002).
70
20. Senorans, F. J. and Ibanez, E. Analysis of Fatty Acids in Foods by Supercritical Fluid Chromatography. Anal. Chim. Acta, 465, 131-144 (2002).
21. Lanser, A. C., List, G. R, Holloway, R K. and Mounts, T. L. FTIR Estimation of Free Fatty Acid Content in Crude Oils Extracted from Damaged Soybeans, J. Am. Oil Chem. Soc., 68, 448-449 (1991).
22. Ismail, A. A., van de Voort, F. R and Sedman, J. Rapid Quantitative Determination of Free Fatty Acids in Fats and Oils by FTIR Spectroscopy. J. Am. Oil Chem. Soc., 70, 335-341 (1993).
23. Bertran, E., Blanco, M., Coello, J., Iturriaga, H., Maspoch, S. and Montoliu, 1. Determination of Olive Oil Free Fatty Acid by Fourier Transform Infrared Spectroscopy, J. Am. Oil Chem. Soc., 76, 611-616 (1999).
24. Man, Y. B., Moh, M. H. and van de Voort, F. R Determination of Free Fatty Acids in Crude PaIm Oil and Refined-Bleached-Deodorized Palm Olein Using Fourier Transform Infrared Spectroscopy, J. Am. Oil Chem. Soc., 76, 485-490 (1999).
25. Verleyen, T., Verhe, R, Cano, A., Huyghebaert, A. and De Greyt, W. Influence of Triacylglycerol Characteristics on the Determination of Free Fatty Acids in Vegetable Oils by Fourier Transform Infrared Spectroscopy, J. Am. Oil Chem. Soc., 78, 981-984 (2001).
26. Zhang, H-Z and Lee, T-C. Rapid Near-Infrared Spectroscopic Method for the Determination of Free Fatty Acid in Fish and Its Application in Fish Quality Assessment. J. Agric. Food Chem., 45, 3515-3521 (1997).
27. Man, Y. B. C. and Moh, M. H. Determination of Free Fatty Acids in Paim Oil by Near-Infrared Reflectance Spectroscopy, J. Am. Oil Chem. Soc., 75, 557-562 (1998).
28. Muik, B., Lendl, B., Molina-Diaz, A. and Ayora-Canada, M. J. Direct, ReagentFree Determination ofFree Fatty Acid Content in Olive Oil and Olives by Fourier Transform Raman Sp~ctrop1etry~ :Anal. Chim. Acta, 487, 211-220 (2003).
29. Canada, M., Medina, A. and Lendl, B. Determination of Free Fatty Acids in Edible Oils by Continuous-Flow Analysis with FT-IR Spectroscopic Detection, Appl. Spectrosc., 55, 356-360 (2001).
71
BRIDGE
In Chapter 4, an approach to FF A analysis was developed based on the treatment
of oils with an oil immiscible solvent containing a weak base to convert the FF A to their
carboxylate salts and concentrate the salts in a small volume to increase the sensitivity of
the method. By comparison with the method described in Chapter 3, this approach also
provided the benefit of eliminating the need to prepare two samples to obtain one
analytical result as well as allowing an overall simplification of the analysis. This
extraction/concentration approach is considered in Chapter 5 as a possible means by
which the moisture content of edible oils can be determined.
72
CHAPTER5
A NEW FTIR METHOD FOR THE DETERMINATION OF MOISTURE IN
EDIBLE OILS
5.1. ABSTRACT
A rapid, practical, and accurate FTIR method for the determination of moisture
content in edible oils has been developed based on the extraction of water from oil
samples into dry acetonitrile. A calibration curve covering a moi sture content range of 0-
2000 ppm was developed by recording the mid-IR spectra of moi sture standards,
prepared by gravimetric addition of water to acetonitrile that had been dried over
molecular sieves, in a 500 J.1m ZnSe transmission flow cell and ratioing these spectra
against that of the dry acetonitrile. Water was measured in the resulting differential
spectra using either the OH stretching (3629 cm-1) or bending (1631 cm-1
) bands to
produce linear standard curves having SDs of - ±20 ppm. For moi sture analysis in oils,
the oil sample was mixed with dry acetonitrile in al: 1 w/v ratio, and after centrifugation
to separate the phases, the spectrum of the upper acetonitrile layer was collected and
ratioed against the spectrum of the dry acetonitrile used for extraction. The method was
validated by standard addition experiments with samples of various oil types, as well as
with oil samples deliberately contaminated with alcohols, hydroperoxides, and free fatty
acids to investigate possible interferences from minor constituents that may be present in
oils and are potentially extractable ,illtO !.acetonitrile. The results of these experiments " ,
confirmed that the moisture content of edible oils can be assessed with high accuracy (on
the order of ±1O ppm) by this method, thus providing an alternative to the conventional,
but problematic Karl Fischer method and facilitating the routine analysis of edible oils for
moi sture content.
5.2. INTRODUCTION
Moisture content is an important parameter associated with the processing and
quality of edible oils. Commonly used in pro cesses such as degumming and refining,
water is subsequently removed by centrifugation, adsorption or vacuum drying. Although
water has limited solubility in fats and oils, the moi sture content must be minimized for
73
the effective adsorption of remaining soap traces after retining and to reduce lipid
hydrolysis during and after processing (1) GeneraIly, the moi sture content ofretined oils
should be <0.1% (1000 ppm) and preferably <500 ppm (2). Methods for moi sture
ap.alysis in edible oils that serve as Official Methods of the American Oil Chemists'
Society (AOCS) (3) include evaporative, distillation or titrimetric procedures as weIl as
the Karl Fischer (KF) method, which is based on the selective reaction of water with a
mixture of iodine, a base, sulfur dioxide, and an alcohol followed by either titrimetric,
potentiometric (4,5) or coulometric (6,7) quantitation. The coulometric KF method is
considered to be the most accurate and sensitive method available and is the gold
standard for the determination of a wide range of moi sture contents (1-25,000 ppm).
However, it is an exacting procedure, uses expensive and environmentally problematic
reagents, and is affected by oil oxidation end products such as aldehydes, ketones, and
hydroperoxides, as these can undergo aldol condensation and/or redox reactions under the
conditions of the test (8). Thus, there is effectively no simple, convenient, sensitive
method available for the analysis of moisture in fats and oils.
At tirst glance, IR spectroscopy would appear to be an obvious instrumental
approach for the direct measurement of moisture in edible oils, given the strong
absorption bands of water in the mid-IR portion of the spectrum, (9) and a limited
feasibility study of the determination of moisture content in crude palm oil by FTIR
spectroscopy has recently been reported (10) However, a number of complicating factors
that were not taken 'into account in: the' previous study must be considered in the
development of a generalized FTIR method for the determination of moi sture in edible
oils. First of aIl, the band shapes and intensities of the water absorptions, particularly the
OH stretching vibrations, are strongly affected by the extent of hydrogen bonding
between water molecules, which will depend on the moisture content in an oil, as weIl as
between water and other hydrogen-bonding constituents. Furthermore, OH-containing
constituents that are frequently present in oils (e.g., alcohols, free fatty acids (FFAs),
and/or hydroperoxides) not only perturb the water absorptions via hydrogen-bonding
interactions but also give rise to spectrally interfering bands. In a recent publication (11),
we reported a new approach to the FTIR analysis of moisture in new and used mineral
based lubricants in whlch similar "matrix effects" were overcome by the combined use of
74
/-the stoichiometric reaction of water with dimethoxypropane (DMP) to produce acetone
(12,13) and differential spectroscopic techniques to quantitate this end product by
measurement of its v (C=O) absorption. This approach, however, is not applicable to
edible oils because the triacylglycerol ester linkages give rise to an intense v (C=O)
absorption that would completely mask the v (C=O) band of acetone in a transmission
spectrum, except at extremely short path lengths. Moreover, the use of ATR to overcome
this limitation is impractical owing to the difficulty of preventing the sample from
picking up moisture from the environment, nor would it provide sufficient sensitivity for
the analysis of moi sture at the low levels present in edible oils. Hence, an alternative
approach is required for edible oils, and this paper describes the evaluation of an
approach based on the extraction of water from the oil into a suitable solvent, culminating
in the development of an acetonitrile extraction method that allows for the accurate
determination of low levels of moisture in edible oils.
5.3. METHODOLOGY
Reagents. HPLC grade acetonitrile, obtained from Fisher Scientific (S~. Louis,
MO), was kept over 4-8 mesh 4A molecular sieves and dispensed using a re-pipette
(Hirschmann-Laborgerate, Germany) protected by desiccant to prevent moisture ingress.
Various refined edible oils were obtained locally, and sorne samples of these oils were
spiked with constituents spectrally representative of contaminants found in edible oils;
these included tert-butyl hydroperoxide, glycerol, and oleic acid, aIl obtained from
Sigma-Aldrich.
Instrumentation. The FTIR spectrometer used for this study was a Bornem
WorkIR (Bornem, Quebec City, PQ, Canada) equipped with a DTGS detector and purged
with dry air using a Balston dryer (Balston, Lexington, MA). The spectrometer was
controlled by an IBM-compatible Pentium 150-MHz PC running proprietary Windows
based UMPIRE® (Universal Method Platform for InfraRed Evaluation) software
(Thermal-Lube, Pointe-Claire, PQ, Canada). A 500 J.lm ZnSe transmission flow cell
(International Crystal Laboratories, Garfield, NJ) equipped with Luer fittings was
employed for the handling of oil samples. As illustrated in Figure 5.1, the cell inlet was
connected to a 10 cm, 18-gauge stainless steel aspiration needle via flexible silicone
75
tubing, and the outlet line was connected to vacuum via a trap and was titted with a valve
to allow both aspiration of the sample through the cell and emptying of the celI; in the
latter process, the air entering the cell was passed through a desiccant tube containing
molecular sieves and Drierite to prevent environmental moisture from contaminating the
analytical system. AlI spectra were collected by co-adding 32 scans at a resolution of 8
cm-1 and a gain of 1.0.
Desiccant
Needle
CellHolder
Valve ~~~o-__ ~Vacuum
--
Acetonitrile layer
Oillayer
Cell Holder Base
Figure 5.1. Schematic diagram of the FTIR sample handling system. To facilitate loading, the upper acetonitrile layer of the centrifuged or separated sample (1) is vacuum aspirated into the IR cell using the 10 cm stainless steel needle attached by flexible tubing to the IR cell. Between sample loadings, the needle is inserted into the desiccant tube outlet (2) to ensurethat air used to flush the tubing and the cell does not introduce any moisture.
Analytical Protocol. The sample preparation procedure and analytical protocol
developed for the determination of the moisture content of oils by extraction of the
moisture into dried acetonitrile and its quantitation by FTIR spectroscopy are illustrated
76
in Figure 5.2. For the analysis of oïl sampi es having moisture contents in the range of 50-
2000 ppm, a 1:1 (v/w) acetonitrile:oil ratio was established to be suitable. Accordingly, 5
g of the sample was added to a tared 15-mL clinical centrifuge tube and weighed on an
analytical balance, with the weight recorded to ±0.001 g, and 5 mL of the dried
acetonitrïle was then added to the tube using a calibrated re-pipette. The tubes were
capped, shaken on a vortex mixer for 30 s, and then centrifuged for 2 min at -6000 rpm
(-5000xg) to separate the oïl and acetonitrile phases. Altematively, adequate separation
of the phases could be achieved by letting the sample stand for -10 min, eliminating the
need for the centrifugation step. Approximately 2 mL of the upper acetonitrile layer was
then aspirated into the transmission flow ceIl, and its spectrum recorded and ratioed
against the spectrum of the dry acetonitrile used to extract moisture from the oïl samples.
A new acetonitrile background spectrum was collected after every 20 samples or after 1
h, whichever came first.
~~ ~ 5 g Acetonitrile
1iI~ -! !
1 Mix 1 ~--+IAJLI !ëentrifugel \t .--------,
l Spectrum 2 - Spectrum 1 = 1 ~ 1
? -+ f~3 --+I~I
Figure 5.2. Schematic diagram of the sample preparation procedure for moi sture analysis of edible oils and the FTIR spectral analytical protocol.
77
Calibration and Validation. Two calibration approaches were evaluated: one based
on the use of primary water/acetonitrile standards and the other on water/oil standards.
Twenty primary water/acetonitrile standards were prepared gravimetrically by adding
distilled water to dry acetonitrile to coyer a range of 0-1000 ppm, and a set of 12
water/oil standards (0-2000 ppm) were similarly prepared by gravimetrically adding
distilled water to canola oïl previously kept over molecular sieves for a minimum of 1
week. The primary standards were analyzed directly by FTIR spectroscopy, while the
water/oïl standards were taken through the analytical protocol illustrated in Figure 5.2. In
both cases, calibration equations were developed relating the moi sture added (in ppm) to
the three measurable water bands at 3629, 3541, and 1631 cm-l, relative to a baseline
point at 2500 cm-l. The reproducibility and accuracy of the FTIR method were assessed
by standard addition of water to various oils (olive, corn, safflower, peanut, and
sunflower), each of these samples being analyzed in triplicate, on different days, with the
FTIR-predicted moi sture contents being compared with the amounts added.
5.4. RESULTS AND DISCUSSION
5.4.1. General Considerations
The objective of this study was to develop a generally applicable method for the
routine analysis of moisture in edible oils based on extraction of the moisture into a
suitable solvent followed by quantitation by FTIR spectroscopy. The solvent initially
considered was 020, in wbich each mqieculeofH20 extracted from the oil sample would "
rapidly undergo hydrogen-deuterium exchange to produce two HOO molecules, thereby
effectively doubling the spectral signal for each molecule of H20 present. Thus, in
princip le, samples could be analyzed by mixing a defined excess of 0 20 with the oil,
allowing the oïl and 020 to separate, and quantitating the HOO formed by measuring the
intensity of the characteristic HOO bands at 3375 and 1450 cm-l in the spectrum of the
0 20 layer. However, although tbis approach worked well for clean refined oils, the
presence of OH-containing constituents in the oil (e.g., alcohols, hydroperoxides or
FFAs), which also undergo hydrogen-deuterium exchange with 0 20, albeit more slowly,
resulted in significant overestimation of moisture content, as evidenced by standard
addition experiments. This problem, in conjunction with the relatively high cost of D20,
78
led us to seek another extraction solvent in which the concentration of H20 could be
accurately quantitated.
Among the solvents investigated, acetonitrile was ultimately selected. It was
found to be sufficiently polar to be immiscible with edible oils and solubilize water while
having limited capacity to solubilize the other potentially interfering OH-containing
constituents that may be present in oils. Acetonitrile is also a highly suitable solvent in
which to measure moi sture levels by IR spectroscopy as it does not absorb strongly in the
portions of the mid-IR spectrum where water absorbs. This lack of spectrally interfering
bands makes it possible to use path lengths of up to 1000 J.lm for the analysis of low
levels of moisture, thereby providing high sensitivity. Finally, in contrast to the spectrum
of water itself, in which the-two OH stretching absorptions are blended into one large
diffuse band owing to the hydrogen-bonding network, the spectrum of water diluted in
acetonitrile at mole fractions of <0.1 exhibits two discrete OH bands, which has been
attributed to the binding of water molecules exclusively to acetonitrile rather than to each
other at these high dilutions (14). Thus, the spectra ofwater added to acetonitrile at levels
of 400 and 800 ppm, presented in Figure 5.3, exhibit clearly delineated bands at 3629
and 3541 cm"l together with the weaker HOH bending vibration at 1631 cm"l.
1.0
1631 cm"I_--a
0.8 CIJ CJ
= 0.6 = ,.Q &. Q rI.l 0.4 ,.Q
-< 0.2
0.0 r i
3600 2000 1600
Wavenumber (cm-1)
Figure 5.3. Spectra ofwater/acetonitrile standards (0, 400, and 800 ppm added water) (upper series) and the corresponding speCtra obtained after subtraction of the spectrum of the acetonitrile employed to prepare the standards (lower series).
79
~'. (
5.4.2. Calibration Development
Primary calibrations were devised using 20 standards prepared by addition of
water to acetonitrile (dried over molecular sieves) to obtain moi sture contents in the range
of 0-1000 ppm based on the gravimetrically added amount. The contribution of any water
present in the acetonitrile was not taken into account in calculating the moi sture contents
of these standards because the method was designed to employ differential spectra,
obtained by subtraction of the spectrum of the acetonitrile used to prepare the standards
from the spectrum of each standard, thereby eliminating the spectral contributions of any
residual water in the dried acetonitrile.
The spectra of the standards were recorded in a 500 Jlm celI, and the following
linear regression equations wère obtained for the 3629, 3541, and 1631 cm-1 bands in the
differential spectra:
H20 (ppm) = 3420.82* A 3629 cm-1 - 4.29
R2 =0.999 SD = 17.7 [5.1]
H20 (ppm) = 3934.92* A 3541 cm-! - 5.55
R2 =0.999 SD = 17.4 [5.2]
H20 (ppm) = 6941.50*A 1631 cm-1 + 2.33
SD = 16.9 [5.3]
These equations indicate relative sensitivities of -3.4, 3.9, and 6.9 ppm per
milliabsorbance unit for the 3629, 3541, and 1631 cm-! bands, respectively, with SDs of
<20ppm.
Figure 5.4 presents a series of differential spectra obtained for a second set of
standards (n = 12) prepared by spiking canola oH with water (0-2000 ppm) and
processing the samples in accordance with the protocol illustrated in Figure 5 2. The
calibration equations derived from these spectra were also linear, with similar slopes,
correlation coefficients, and regression SDs as those obtained using the primary
80
r-..
water/acetonitrile standards, but had negative intercepts of ~280 ppm, indicating that the
canola oil contained about 0.03% moisture. A notable difference between the differential
spectra of the two sets of standards (Figure 5.3 vs. Figure 5.4) is the presence of a
carbonyl band (peak D) at 1740 cm- l in the spectra of the set prepared by extraction of
water from oil into acetonitrile. It was detennined experimentally that the positions of the
v (C=O) absorptions of oils (triacylglycerol ester linkage) and FFA mixtures derived from
the same oils were identical in acetonitrile. However, although FFAs are freely soluble in
acetonitrile, it was considered unlikely that peak D was solely due to FF As extracted
from the oïl, owing to their low levels in refined oils «0.05%), and this was borne out by
further examination of the spectra, in that peak D was not accompanied by any FF A
absorption at ...... 3300 cm- l. Thus, the observation ofpeak D was taken as an indication that
sorne amount of oil was extracted into the acetonitrile layer.
Figure 5.4. Differentiai spectra obtained after extraction of water from water/canola oil standards (0-2000 ppm water) into acetonitrile, recording the spectrum of the acetonitrile layer, and ratioing the spectrum against that of the acetonitrile extraction solvent. The major spectral features are identified as follows: A and B (3629 and 3541 cm-l, respectively), OH stretching vibrations; C, water association band; D (1740 cm- l
),
ester carbonyl band of oil extracted into acetonitrile; E (1631 cm-\ HOH bending vibration.
81
Accordingly, the miscibility of various oils with acetonitrile was assessed by
mixing the oil with acetonitrile in al: 1 (w/v) ratio and measuring the height of the
carbonyl band in the acetonitrile spectrum after phase separation was complete. The oils
were detennined to be slightly miscible (-0.5%) to an extent that varied somewhat with
the overall degree of unsaturation, which could mean that moi sture detennination using
the primary (water/acetonitrile) calibration could be affected by a dilution error that
would depend on the type of oil analyzed. To investigate this possibility, five oils (olive,
corn, safflower, peanut, and sunflower), all pre-dried over molecular sieves, were each
spiked with three levels ofmoisture (approximately 100, 300, and 500 ppm) and analyzed
in duplicate in accordance with the analytical protocol depicted in Figure 5.4. Regression
(forced through the origin) of the means of the duplicates against the amount of moisture
gravimetrically added to all the oils produced the following Z-reg relationship:
Predicted H20 = 0.986 * Added H20
R2 = 0.998 SD = 8.1 ppm [5.4]
Based on this experimental evidence, the use of calibrations based on simple
water/acetonitrile standards was considered a suitable means by which to quantify
moisture in edible oils.
5.4.3. Consideration of Interfering Con$tituents
Many minor constituents that may be present· in oils can be extracted into
acetonitrile and hence are potential sources of spectral interference in the quantitation of
moisture by the method developed in the present study. Of particular concern are OH
containing molecules such as FF As, hydroperoxides, and alcohols (glycerol, mono- and
diglycerides). Indeed, as illustrated in Figure 5.5, which shows the overlaid differential
spectra of acetonitrile spiked with glycerol, hexanol, tert-butyl hydroperoxide, oleic acid,
and water, most of these constituents have bands that overlap significantly with one of the
two OH stretching bands (3541 cm-1) employed for quantitation of moisture. However,
the higher-frequency OH stretching band (3629 cm-1) is largely in the clear.
82
Although the HOH bending band (at 1631 cm-1) does not suffer from
interferences from OH-containing constituents, the use of this band for quantitation of
moi sture is not optimal because of the resulting ~2-fold reduction in sensitivity (cf. Eqs.
[1]-[3]).
A 0.6
0.4
0.2
0.0 ....... --
3600 3400 3200 3000
Wavenumber {cm-Il
,t,
" . Figure 5.5. Differentiai spectra of acetonitrile spiked with hexanol (A), glycerol (B), tert-butyl hydroperoxide (C), oleic acid (0), and water (shaded). Note: The spectra are not on the same scale.
The effects that OH-containing constituents present in an oil would have on the
accuracy of the method was evaluated quantitatively by analyzing 24 standards prepared
gravimetrically by spiking corn oil with water (0-1000 ppm) and further spiking portions
of each of these standards with (i) oleic acid (0.09-0.8% w/w), (ii) glycerol (0.05-0.1 %
w/w) , or (iii) tert-butyl hydroperoxide (0.1-1 % w/w). Comparison of the moisture
predictions for the 18 standards in subsets (i), (ii), and (iii) with those for the 6 standards
spiked only with water demonstrated that the predicted values obtained using the 3541
83
cm-1 band were high, roughly in proportion to the amount of "contaminant" added; the
largest effect was observed for glycerol, as would be expected given the extensive band
overlap illustrated in Figure 5.5, followed by the hydroperoxide and then the FF A. On
the other hand, the predictions of moi sture content based on either the 3629 or the 1631
cm-1 band showed no such effects. Regression of the predicted moi sture contents for an
additional subset, prepared by spiking four water/oil standards with randomized amounts
of all three contaminants, against the "expected" values (calculated from the predictions
for the standards by correction for the dilution due to spiking) yielded the following Z-reg
equations for the 3629 and 1631 cm-1 bands, respectively:
Predicted H20 3629 cm-1
= 1.054 * "Expected" H20
R2 =0.998 SD = 11.9ppm [5.5]
Predicted H20 1631 cm-1 = 1.050 * "Expected" H20
R2 =0.997 SD = 18.3 ppm [5.6]
These results éonfirmed that peak height measurements at 3629 or 1631 cm-1
allowed for the accurate quantitation ofmoisture in oils in the presence ofOH-containing
constituents at the levels commonly associated with edible oils.
5.4.4. Factors Affecting A~alytical Accu'tacy
Owing to the ubiquitous nature of moisture in the environment, precautions were
required to avoid moisture contamination of the sample handling system A simple, but
effective means of minimizing moi sture contamination was devised, as described in the
Experimental section and depicted in Figure 51, without which reproducible calibrations
over the range of 0-200 ppm could not be obtained. Above 200 ppm, the effects of
environmental moisture were relatively minor; however, to cover aIl contingencies, this
sample handling system was used routinely. Furthermore, in experiments conducted to
determine the relative degree of analytical stability ànd reproducibility when samples
were analyzed over a period of three days, deterioration of the analytical performance as
. time progressed was observed and eventually attributed to slow adsorption of water on
84
the walls of glass centrifuge tubes. This was confirmed by the finding that reproducible
results were obtained in similar experiments when plastic centrifuge tubes fitted with
septa were employed in place of capped glass tubes. The results of these experiments not
only convey sorne of the precautions that must be taken to prevent loss or pickup of )
moi sture when analyzing low levels of moisture in oil samples but also serve to illustrate
the sensitivity of the FTIR method.
With proper precautions, very accurate data (on the order of ±10 ppm) can be
obtained, based on standard addition experiments. For maximum sensitivity, the 3629 cmc
1 OH stretch is the preferred measurement; however, the 1631 cm-l band is useful for
higher moi sture levels (above 2000 ppm). Although not detailed in this paper, additional
sensitivity (down to ~2 ppm) can be attained by increasing the pathlength (upper limit
~ 1 000 ~m) or using a 2: 1 oil:acetonitrile ratio in the extraction step, or both. However, at
such low moisture levels, the main limitation is not the sensitivity of the spectroscopic
measurement per se, but rather the difficulty of ensuring against moi sture pickup or loss,
especially after extraction and during analysis.
5.5. ACKNOWLEDGEMENTS
The authors would like to thank Sultan Qaboos University, Oman, and the Natural
Sciences and Engineering Research Council (NSERC) of Canada, for their financial
support of this research. Thanks to Charles Chaney for suggestions related to this work
and to Sophie Archelas for her technical assistance.
5.6. REFERENCES
1. O'Brien, R. D., Fats and Oils: Formulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998.
2. Rossell, J. B. "Classical Analysis of Oils and Fats", in Analysis of Oils and Fats, R. J. Hamilton and J. B. Rossell, Eds., Elsevier Applied Science Publisher Ltd., London, England, 1986, Chap. 1, p. 52.
3. Official Methods and Recommended Practices of the American Oil Chemists' Society (AOCS, Champaign, IL, 1989), 4th edn.
4. Cedergren, A. and Luan, L. Potentiometric Determination of Water Using Spent Imidazole-Buffered Karl Fischer Reagents. Anal. Chem., 70, 2174-2180 (1998).
85
5. Rosvall, M., Lundmark, Land Cedergren, A. Computer-Controlled, Coulometric Karl Fischer System for Continuous Titration of Water Based on Zero CUITent Potentiometry. Anal. Chem., 70, 5332-5338 (1998).
6. Cedergren, A and Jonsson, S. Diaphragm-Free Cell for Trace Determination of Water Based on the Karl Fischer Reaction using Continuous Coulometric Titration. Anal. Chem., 69, 3100-3108 (1997).
7. Cedergren, A. and Jonsson, S. Progress in Karl Fischer Coulometry using Diaphragm-Free Cells. Anal. Chem., 73, 5611-5615 (2001).
8. Cedergren, A. Reaction Rates between Water and the Karl Fischer Reagent. Ta/anta, 25, 256-271 (1978).
9. Smith, B. Infrared Spectral Interpretation: A Systematic Approach (CRC Press, Boca Raton, FL, 1999), Chap. 1, p. 22.
10. Man, Y. B. C and Mirghani, M. E. S. Rapid Method for Determining Moisture content in Crude Palm Oil by Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 77, 631-637 (2000).
11. van de Voort, F. R., Sedman, J., Yaylayan, V., Saint Laurent, C. and Mucciardi, C. Quantitative Determination of Moisture in Lubricants by Fourier Transform Infrared Spectroscopy. Appl. Spectrosc., 58, 193-198 (2004).
12. Critchfield, F. E. and Bishop, E. T. Water Determination by Reaction with 2, 2-Dimethoxypropane. Anal. Chem., 33, 1034-1035 (1961).
13. Dix, K. D., Sakkinen, P. A., and Fritz, J. S. Gas Chromatographie Determination of Water Using 2, 2-Dimethoxypropane and a Solid Acid Catalyst. Anal. Chem., 61, 1325-1327 (1989).
14. Le Narvor, A., Gentric, E. and Saumagne, P. Study of Binary Mixtures of Water with Severa! Proton Acceptors by Infrared Spectroscopy in the Region of OH Stretching Bands. Cano J. Chem., 49, 1933-1939 (1971).
86
BRIDGE
The methods developed in Chapters 4 and 5 share a common analytical approach,
namely the extraction of a constituent of interest from an edible oU into an oU immiscible
solvent and its spectroscopic quantification. The development of these sensitive and
accurate methods is only part of the solution in an industry that is looking to maximize
efficiency and reduce costs through automation. Because the extraction based procedures
use low viscosity solvents, they are amenable to being implemented on an automated
system equipped with an auto-sampIero Chapter 6 describes the implementation of
algorithms and procedures as well as validation of the methodology developed to effect
the automated analysis of FF A and moi sture in edible oils using such a system based on
the methods described in Chapters 4 and 5.
" ,
87
CHAPTER6
AUTOMATED FTIR ANALYSIS OF FREE FATTY ACIDS OR MOISTURE IN
EDIBLEOILS
6.1. ABSTRACT
An FTIR spectrometer coupled to an auto-sampler and attendant methodologies
for high-volume automated quantitative analysis of free fatty acids (FF A) and moi sture in
edible oils are described. Samples are prepared by adding 20 g of oil to a 50 ml screw
capped vial, to which Îs added either a methanol/NaHNCN solution or dry acetonitrile in
a 1: 1 (w/v) ratio for FF A or H20 analysis, respectively. After capping with Mylar-lined
septum caps, the vials are loaded into an auto-sampler tray, which is then agitated
vigorously to extract the constituent of interest from the oil into the oil-immiscible
solvent, and are then left to stand for ~ 10 min to allow for phase separation. The upper
solvent layer in each vial is aspirated successively into the IR ceIl, with the Mylar seal
allowing facile auto-sampler needle penetration into the vials. The spectra of the sample
extraction solvents serve as spectral backgrounds as weIl being used to monitor cell path
length and verify cell loading. FF A and H20 analyses are carried out using 100 /-lm and
500 /-lm CaF2 ceIls, respectively. For FFA analysis, quantification is achieved using the
v (COO-) band at 1573 cm-l, while moisture is determined using water absorption bands
at 3541 cm-l or 1631 cm-l, depending on the moisture range of the samples. Calibration
procedures and data are presented. The spectrometer and auto-sampler are controlled
using proprietary UMPlRE Pro® (Universal Method Platform for Infra Red Evaluation)
software, which provides a simple user interface and automates the spectral analysis; the
output data can also be sent to a Laboratory Information Management System (UMS).
Validation and performance data obtained with this automated system demonstrate that it
is capable of analyzing >60 samples/h, a rate commensurate with the throughput required
by commercial contract or high-volume process controllaboratories.
6~2. INTRODUCTION
Edible oils represent one of the primary constituents required for the formulation
and manufacture ofproducts by the food industry. Oils are extracted from a wide range of
88
raw materials and generally undergo relatively standardized extraction and refining
processes and supplemental transformations such as fractionation, hydrogenation and/or
interesterification prior to use. During processing, a variety of physico-chemical oïl
quality parameters are monitored, including the determination of free fatty acids (FF A),
moisture (H20), peroxide value, iodine value, and saponification number, to name just a
few1• Typically, QC laboratories have been limited to using standardized (AOCS2 or
IUP AC3) manual wet chemical methods to monitor oïls in process, requiring a variety of
analytical setups and multiple reagents as well as skilled personnel to carry them out,
although these methods have been automated to sorne extent through the use of automatic
titrators and flow injection systems.
The Mc Gill IR Group has worked toward developing Fourier transform infrared
(FTIR) spectroscopy as a rapid instrumental technique for the routine analysis of edible
oils. Methodologies have been deve10ped for the determination of FF A 4,5, saponification
number6, iodine value6
,7, trans and cis content',9, ansidine valuelO, solid fat index 11 , and
peroxide valueI2-14
, amongst others. From the standpoint of implementation, most of
these methods are semi-automated, with the software providing a simple user interface as
well controlling the instrument, performing all calculations, and storing the results
automatically. These FTIR methods provide improved sample turnaround and a reduction
in reagent use relative to wet chemical methods. However, they do not provide the leve1
of speed required by contract or centralized QC laboratories handling more than 100
samples/d, the key factor limiting sample throughput being the high viscosity of edible
oils. On the other hand, in the case of new FTIR methods for the determin'ation of FF A 15
and H2016 in edible oils that have recently been deve10ped in our laboratory, this
limitation is obviated because the constituent of interest is extracted into a low-viscosity
solvent prior to FTIR analysis. Indeed, these new FTIR methods have the potential for
much higher sample throughputs than auto-titration systems, which presently provide the
most rapid means of obtaining reliable FF A or moisture data. In addition to reducing the
sample-flow constraints associated with high-viscosity oils, these extraction-based
methods largely e1iminate problematic matrix effects and provide much greater
sensitivity than can generally be achieved by direct FTIR analysis of oils. This paper
describes the instrumentation, analytical protocols, and software associated with an
89
automated FTIRIauto-sampler system designed to analyze for either FF A or H20 in
edible oils at rates ofup to 100 samples per hour.
6.3. METHODOLOGY
Instrumentation. The analytical system used couples a Bornem WorkIR (Bornem,
PQ, Canada) FTIR spectrometer with a Model 223 Gilson auto-sampler (Gilson, Inc.,
Middleton, WI) and a positive displacement micro-pump (Vissers Sales Corp., ,Aurora,
Ont., Canada). The spectrometer is equipped with a DTGS detector and uses 100 and 500
~m CaF2 transmission flow cells (International Crystal Laboratories, Garfield, NJ, USA)
for FF A and H20 analysis, respectively, with Luer-Lock cell connectors used to facilitate
cell changes.
Software. The spectrometer, micro-pump and auto-sampler were all controlled by
an IBM-compatible PC running proprietary Windows-based UMPlRE® Pro (Universal
Method Platform for InfraRed Evaluation) software, developed by Thermal-Lube Inc.
(Pointe-Claire, PQ, Canada). Built on the Microsoft.NET Framework, UMPlRED Pro
software is designed to work with Windows XP or Windows 2000 operating systems.
UMPlRE Pro is a self-contained platform that allows for the acquisition of spectra,
analysis of spectral data, and cataloguing of results without operator intervention. For
systems equipped with an autosampler and pump (such as that used in the present work),
the analysis of all the samples loaded into the autosampler tray is fully automated. All
parameters àssociated with the three fundamental tasks controlled by the software
(spectral acquisition, data processing, and cataloguing of results) are predefined for each
type of analysis (e.g., FF A determination, moi sture determination) in an analysis method;
additional customized analysis methods can be created using the method development
tool provided in the software. The analysis method specifies all relevant acquisition
parameters relating to the spectrometer and the auto-sampler, the component(s) to be
quantified, and the component method associated with each component, where the latter
specifies the spectral features that will be used for its quantification and the calibration
(linear or polynomial) by which the results will be computed. AlI results are archived
within the software and can be retrieved by referencing either the auto-sampler tray,
analysis method, operator that acquired the sampI es, value of result, etc. These results can
90
be printed, exported to external software for further analysis or reporting, or re-processed
by a different analysis method. Through the use of eXtensible Markup Language (XML)
and the Microsoft.NET environment, UMPIRE Pro can relay information to a Laboratory
Information Management System (UMS).
Reagents. AlI solvents were of HPLC grade, obtained from Fisher Scientific (St.
Louis, MO). Acetonitrile was dried over 4-8 mesh 4A molecular sieves for a minimum of
two weeks and dispensed using a re-pipette (Hirschmann-Laborgerate, Germany)
protected by Drierite to prevent moisture ingress l6. The reagent solution employed in the
FFA analysis was prepared by dissolving NaHNCN in MeOH (2 g/L). This solution was
allowed to stand for ~4 days or until the v (C=N) band at 2100 cm- l completely
disappeared before use l5. Both extraction solvents (acetonitrile and methanol/NaHNCN)
were kept in sealed bottles and maintained over molecular sieves.
Calibration Standards. For the FF A calibration, calibration standards were
prepared by adding palmitic acid (0-1 %) to canola oil which had been passed through
activated silica gel column to remove any traces of FF A and other oxygenated
compounds. Moisture calibration standards were prepared by gravimetric addition of
distilled water (0-1000 ppm) to corn oil which had previously been kept over 10-20 mesh
Drierite (calcium sulfate) for at least one week.
Analytical Protocol. Both the FF A and moisture analyses are based on extraction
of the constituent of interest into an oil-immiscible solvent. For the analysis of FF A, the
extraction uses methanol containing a weak base, hydrogen cyanamide (NaHNCN), the
latter producing immediate conversion of acids into methanol-soluble salts without
causing saponification of the oils, whereas for moisture analysis the sample is extracted
with dry acetonitrile. The details associated with the procedures for these analyses have
been described elsewhereI5,16. Figure 6.1 presents a generalized schematic diagram of the
analytical proto col folIowed for both the calibration standards and the samples. For either
FF A or moi sture analyses, 20 g of oil was weighed into a tared 50-mL plastic vial to
within --±0.1 0 g. To the vial containing oil, 20 ml of the appropriate solvent was added
using a calibrated pro-pipette, and the vial was capped. The vials were agitated on a
vortex mixer for 30 s, and then loaded onto the autosampler tray. The thoroughly mixed
91
/~~
samples were then left to stand for 10-15 min to ensure separation of the oil and solvent
layers. The 'automated FTIR spectral collection procedure was initiated with the
collection of an open-beam spectrum followed by a single-beam solvent spectrum, the
latter ratioed against the former to produce a reference solvent spectrum used to
determine cell path length. The single-beam spectrum of the solvent layer in each sample
vial was then recorded and ratioed against the single-beam solvent spectrum to obtain the
spectrum of the constituent being analyzed for; the same single-beam solvent spectrum
was used for a complete auto-sampler tray (56 samples). AlI spectra were collected by co
adding 32 scans at a resolution of 8 cm-1 and a gain of 1.0.
20g Oil ..
20ml .. Solvent
! j ( Cap and ) Mix
t Spectrum 1 ~ [JlJL] (Background)
Let """ Separate
(----Â- ) i (Spectrum 2- Spectrum 1)
./ Upper
Spectrum 2 · (JuiL) -Layer (Sample)
Figure 6.1. General analytical protocol for FF A and H20 analysis of edible oils by FTIR spectroscopy.
For FFA analysis, the absorbance of the v (COO-) band at 1573 cm-1 relative to a
baseline of 1820 cm-1 was measured and the FF A content predicted from a calibration
equation obtained by regressing the corresponding absorbance values of the FF A
calibration standards against their FF A content, expressed as % FF A (oleic acid). For
moi sture analysis, the calibration equations used were based on relating the amount of
92
moisture added to the moi sture standards to the absorbance of the strong v (O-H) band at
3629 cm-l and/or the weaker v (H-O-H) band at 1631 cm-l, both measured relative to a
single-point baseline at 2500 cm-l. We have established empirically that measurement of
the more sensitive 3629 cm-l band is optimal up to a moisture level of 2000 ppm, beyond
which the 1631 cm- l band is preferred. The 2000 ppm limit corresponds to an absorbance
at 3629 cm- l of ~0.7, and the switchover from the 3629 cm- l to the 1631 cm- l calibration
equation is automatically made by the software based on this absorbance criterion.
System Evaluation and Validation. Optimization of the analnical·system in terms
ofminimizing sample carryover was initially evaluated by sequentially analyzing low (L)
and high (H) FF A or moisture standards and calculating percent carryover as [(L2 -
L1)/H1] * 100. Sample carryover was further assessed by analyzing three replicates of six
different oils of unknown FF A or H20 content, these 18 samples being placed in the
autosampler tray in random order. Subsequently, reproducibility and accuracy of the
optimized automated FTIR method were assessed by standard addition of palmitic acid or
water to olive, sunflower, and peanut oils, each validation sample being analyzed in
duplicate, on different days, by the appropriate FTIR method and the results compared to
the gravimetrically added amounts of these constituents.
6.4. RESULTS AND DISCUSSION
6.4.1. System Overview and Software
Figure 6.2 illustrates the overall system configuration. Because the FF A and
moisture methods are designed to analyze low-viscosity solvent extracts rather than oils
per se, a small micro-pump can be used to aspirate the sample into the IR cell. Low dead
volume combined with 1/8" i.d. tubing minimizes sample volume (~15 ml required) and
allows the elimination of solvent rinses between sampI es. The sample vials have
conventional plastic septum caps but the silicone septum is replaced by a 20 mil Mylar
liner. This Mylar liner provides the hermetic seal required for moi sture analysis while
being readily punched through by the auto-sampler needle (Figure 6.3).
Both the FTIR spectrometer and the auto-sampler are controlled by UMPIRE Pro,
an upgrade of a proprietary software package designed as a general platform for
developing and implementing FTIR methods. Details regarding the design of the software
93
E
D
Figure 6.2. The FTIRIauto-sampler system used for the analysis of FF A or moisture. The system is composed of micro pump (A), cell holder (B), FTIR spectrometer (C), autosampler tray (D) and robotic arm (E).
Figure 6.3. Auto-sampler needle punching through the Mylar liner.
94
are presented in the Methodology section, and sorne features of the user interface are
illustrated in the panels of Figure 6.4. The user interface consists of a nested set of tab
based menus that allow the user to implement calibration and method development
procedures. Figure 6.4A presents the "General" descriptive tab associated with the FF A
method, and Figure 6.4B presents the Scanning/Sampling tab, from which the FTIR
operating parameters and auto-sampler/pump parameters can be input. Figure 6.4C
illustrates the typical tabs included under the "Components" submenu, showing the tabs
that can be accessed from this submenu to define the peak height or area measurements to
Figure 6.4. The user interface of UMPlRE® Pro program. A and B, the "General" and "Scanning/Sampling" tabs for the "Modify Methods" submenu, respectively; C, "Peak! Area Position" tab for the "Components" submenu.
95
be used for quantification and input calibration equations, waming messages for the
operator, altemate equations to be used under defined conditions, and baseline correction
procedures. This tab-based user interface makes it simple and intuitive to develop, assess,
and modify methods as well as calibration procedures. In addition to these development
and setup functions, spectra can be viewed, manipulated, and exported, and the analytical
result(s) can be displayed, printed, trended and/or associated with a client, machine or
operation as well as output to a Laboratory Information Management System (UMS) for
further processing or reporting.
For the present work, the autosampler was equipped with a 56-slot (7x8) tray. The
first two slots of the auto-sampler tray were reserved for the solvent used for the analysis
in question, the spectrum of which not only served as a background spectrum but also
was employed to determine the cell path length. For the latter purpose, a path length
calibration equation was developed by building three cells of different path lengths and
relating their path lengths, as determined from the fringe count in the spectrum of the
empty cell, to the height of an on-scale solvent band. Typical linear regression equations
obtained for acetonitrile and methanol, respectivelY' were:
Pathlength (/lm) = 673.51 * A 2627/2580 cm-1
- 18.488
R2 = 0.999 SD = 1.75 [6.1]
Pathlength (/lm) = 261.99* A 2044/1952 cm-1
- 0.251
R2 = 0.999 SD = 0.384 [6.2]
Using these equations, a cell path length is determined at the start of each run, of
which an ongoing record is maintained to monitor the condition of the cell over time.
This path length value is employed to normalize the spectral data subsequently collected
for the samples in the tray to a constant path length (500 and 100 /lm for H20 and FFA,
respectively). This procedure also allows one to rebuild cells without performing a
recalibration of the method.
96
Once the auto-sampler tray is loaded with samples in capped vials, the whole tray
is shaken vigorously to extract the constituent of interest from the oil into the solvent and
then the samples are left to stand until phase separation occurs (~1 0 min). Automated
analysis of the samples is then initiated. For each successive sample in the tray, the
autosampler needle punches through the Mylar liner of the cap into the upper solvent
layer, which is then aspirated into the IR cell and its spectrum recorded. A "cell-full"
check is performed on the spectrum by measuring the height of a suitable solvent band in
a spectral region in which the component of interest does not absorb (generally, the same
band as used for the path length calculation); ifthis height is not within specifications, the
analysis is paused, and the operator is notified to check for air bubbles in the cell or
incomplete cellioading. When the "cell-full" check passes, the value of the parameter of
interest is automatically calculated from the calibration equation defined for the method.
Figure 6.5 and Figure 6.6 illustrate the differential spectra (solvent subtracted
out) obtained for moisture and FF A standards, respectively, after extraction with the
appropriate solvent. The calibration equations developed using these spectra were as
follows:
H20 (ppm) = 3418.70*A 3629cm-1 - 61.89
R2 = 0.998 SD=23.8 [6.3]
FFA (% w/w) = 5.05 * A 1573 cm-1 - 0.008
R2 = 0.999 SD = 0.0018 [6.4]
These calibration equations were obtained using calibration standards prepared by
spiking oils with the constituent of interest and extracting them with the appropriate
solvent in accordance with the analytical protocol employed for sample analysis.
However, since the spectral analysis is performed on the upper solvent layer containing
the extracted constituent, the matrix effects associated with the oils are effectively
eliminated, and hence the calibration procedure could be simplified substantially by
spiking the constituents directly into their respective solvents.
97
A 0.3
0.2 ~ ~ = ~ .c ""' 0 <Il 0.1 .c ~
3600 i
2000
Wavenumber (cm-1)
C
D
i
1600
Figure 6.5. DifferentiaI spectra of canola oil standards spiked with water (0-2000 ppm) after extraction with acetonitrile. The major spectral features are the strong O-H stretching vibrations @ 3629 (A) and 3541 cm-1 (B), respectively, the ester carbonyl band due to traces oftriacylglycerol being extracted into acetonitrile @ 1740 cm-1 (C), and the weaker HOH bending vibration ofwater @ 1631 cm-1 (D).
A
0.2
0.00 -1----
1800 1700 1600 1500
Wavenumber (cm-1)
Figure 6.6. DifferentiaI spectra of canola oil standards spiked with palmitic acid (0.0-0.1 % w/w) after extraction with methanol containing NaHNCN. The spectra illustrate the response of the FFA salts produced which absorb @ 1573 cm-1 (A).
98
'~',
6.4.2. System Evaluation and Validation
In optimizing the system parameters, the crucial variable is the pump time, the
objective being to minimize sample turnaround time white avoiding sample carryover.
The optimum pump time is largely a function of solvent viscosity and the wash volume
required to flush out the previous sample. Preliminary trials for setting the pump time
simply involved the analysis of high, intermediate and low FF A or moisture standards
placed in the auto-sampler tray in random order. Based on this screening work, it was
found that consistent results, without any carryover bias, were obtained using pump times
on the order of 30 s( corresponding to a sample volume of ~ 15 ml). with negligible
carryover «0.5%). Subsequent to these trials, a series of three replicate sets of six oils
(olive, sunflower, canola, sesame, grape seed, and safflower) of unknown FF A and
moisture content were prepared for analysis. These 18 samples were loaded into the auto
sampler tray in random order for analysis by each method. Table 6.1 presents the
analytical data obtained for these runs. The excellent SD values confirm that sample
carryover was negligible.
Table 6.1. analyses)
FFA and Moisture Content of Various Edible Oils (mean of triplicate
OH
Canola
Extra virgin olive
Sunflower
Safflower
Sesame
Grape seed
%
0.019
0.270
0.024
0.024
1.170
0.025
FFA
sn 2.9*10-3
10* 10-3
2.9*10-3
5.8*10-3
2.6*10-3
10* 10-3
Moisture Content
ppm sn 39 1.9
850 12.3
145 9.8
87 5.2
302 1.9
168 12.0
Subsequently, a series of validation samples were prepared by gravimetrically
spiking refined olive, corn, peanut, and sunflower oils with random, but known amounts
ofwater (100-1000 ppm) and FFA (0.1-1%). These samples were analyzed in duplicate
using the appropriate FTIR method for each constituent. The FTIR data are plotted
against the amounts of moisture or FF A added to these oils in Figure 6.7 and Figure 6.8,
99
<~,
respectively, and the corresponding zero-regression equations (forced through the origin)
are presented in Eqs. [6.5] and [6.6].
R2 = 0.997 SD = 14 ppm [6.5]
FTIR FFA = 1.011 FFA
R2 = 0.999 SD = 0.007 % FFA [6.6]
--"0 1.2 •• <J
* CI: ~ •• 1.0 QI -0
"t. 0.8 i ~ ~
.-0.6
~ .-
~ 0.4 "0 QI • == •• f e 0.2 :.. •• Qi .... QI
0.0 ; ~
0.0 0.2 0.4 0.6 0.8 1.0
Added FFA (w/w%, Oleic acid)
Figure 6.7. Plots for moi sture content vs. amounts added for the validation run. Error bars represent mean ± SD of three replicates.
100
/..------.....
~,
700
~ 600 ~ e ! Q.
! Q. 500 '-"
0 i N 400
== !
"0 ! ~ 300 == .~
e 1- 200 1 ! ~ ; ~ 100
0 0 100 200 300 400 500 600
Added H20 (ppm)
Figure 6.8. Plots for FF A vs. amounts added for the validation run. Error bars represent mean ± SD of three replicates.
6.5. CONCLUSION
This study providesevidence that the automate d, high-speed analysis of edible
oils for FF A and moisture using a FTIR spectrometer coupled to an auto-sampler is
workable. Following sample preparation, which involves a simple in-vial extraction of
the constituent in question into a low-viscosity solvent, and loading of samples onto an
auto-sampler tray, the analysis is fully automated, with the FF A or moisture content
results being output on-screen or to a LIMS. When optimized, throughputs of ~ 100
samples/h can be achieved, which meets the requirements of high-volume contract,
process, or payment laboratories. It is also likely that the instrument configuration
described in this paper can be adapted to other FTIR-based methods such as the
determination of peroxide value or trans content.
101
6.6. ACKNOWLEDGMENT
The authors would like to thank Sultan Qaboos University, Oman, and Thermal
Lube Inc., Pointe-Claire, PQ, Canada for their financial support of this research. Thanks
also go to Dr. R Cocciardi for his technical assistance.
6.7. REFERENCES
1. O'Brien, RD., Fats and Oils: Formulating and Processing for Applications, Technomic Publishing Company, Lancaster, PA, 1998.
2. AOCS, Official Methods and Recommended Practices of the American Oil Chemists' Society, 4th edn., AOCS Press, Champaign, IL, 1998.
3. IUPAC, Standard Methodsfor the Analysis ofOils, Fats and Derivatives, 7th edn., Blackwell Scientific, London, 1992.
4. Ismail, A. A., van de Voort, F. R. and Sedman, J. Rapid Quantitative Determination of Free Fatty Acids in Fats and Oils by FTIR Spectroscopy. J. Am. Oil Chem. Soc., 70, 335-341 (1993).
5. AI-Alawi, A, van de Voort, F. R and Sedman, J. New Method for the Quantitative Determination of Free Fatty Acids in Oil by FTIR Spectroscopy. J. Am. Oil Chem. Soc., 81,441-446 (2004).
6. Li, H., van de Voort, F. R, Sedman, J. and Ismail, A. A. Rapid Determination of Cis and Trans Content, Iodine Value, and Saponification Number of Edible Oils by Fourier Transform Near-Infrared Spectroscopy. J. Am. Oil Chem. Soc., 76, 491-497 (1999).
7. van de Voort, F. R, Ismail, A. A. and Sedman, J. A Rapid, Automated Method for the Determination of Cis and Trans Content of Fats and Oils by Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 72, 873-880 (1995).
8. Li, Hui; van de Voort, F. R, Ismail, A. A., Sedman, J. and Cox, R Trans Determination of Edible Oils by Fourier Transform Near-Infrared Spectroscopy. J. Am. Oil Chem. Soc., 77, 1061-1067 (2000).
9. Sedman, J., van De Voort, F. R and Ismail, A. A. Simultaneous Determination of Iodine Value and Trans Content of Fats and Oils by Single-Bounce Horizontal Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 77, 399-403 (2000).
10. Dubois, J., van de Voort, F. R, Sedman, J., Ismail, A. A. and Ramaswamy, H. R Quantitative Fourier Transform Infrared Analysis for Anisidine Value and Aldehydes in Thermally Stressed Oils. J. Am. Oil Chem. Soc., 73, 787-794(1996).
11. van de Voort, F. R, Memon, K. P., Sedman, J. and Ismail, A. A. Determination of Solid Fat Index by Fourier-Transform Infrared Spectroscopy. J. Am. Oil Chem. Soc., 73, 411-416 (1996).
102
~--/
12. van de Voort, F. R., Ismail, A. A., Sedman, J., Dubois, J and Nicodemo, T. The Determination of Peroxide Value by Fourier Transform Infrared (FTIR) Spectroscopy. J. Am. Di! Chem. Soc., 71 921-926 (1994).
13. Ma, K., Van De Voort, F. R., Sedman, J. and Ismail, A. A. Stoichiometric Determination of Hydroperoxides in Fats and Oils by Fourier Transform Infrared Spectroscopy. J. Am. Di! Chem. Soc., 74, 897-906 (1997).
14. Li, H., Van de Voort, F. R., Ismail, A. A. and Cox, R. Determination ofPeroxide Value by Fourier Transform Near-Infrared Spectroscopy. J. Am. Di! Chem. Soc., 77, 137-142 (2000).
15. Al-Alawi, A., van de Voort, F. R. and Sedman, J. A New FTIR Method for the Analysis ofLow Levels ofFFA in Refined Edible Oils. Spectrosc. Lett. 38,389-403 (2005).
16. Al-Alawi, A., van de Voort, F. R. and Sedman, J. A New FTIR Method for the Determination of Low Levels of Moisture in Edible Oils. Appl. Spectrosc., (in press, 2005).
103
CHAPTER 7
GENERAL CONCLUSION
The methods documented in this thesis add a significant portion to the continuing
work of FTIR methodology development in oil analysis. Each part of the work laid out in
this thesis provides a solution to a cornrnon problem in its field or presents an improved
me ans of carrying out edible oil analysis. The first method described in Chapter 3
provides a novel and yet workable approach to overcome matrix effects which for years
hindered development of a general method for free fatty acids analysis of edible oils. The
method uses a suspension of a weak base, potassium phthalimide (K-phthal) in 1-
propanol (l-PrOH), to convert FF As present in oils to their carboxylate salts without
causing oil saponification. The matrix effects were resolved by splitting the diluted oil
samples into two halves, with one half treated with the K-phthal reagent and the other
halfwith I-PrOH (blank reagent). The spectra of the two halves were then coUected, and
a differential spectrurn obtained to ratio out the invariant spectral contributions from the
oil sarnple. Quantification of FF A was based on direct measurement of the peak height of
the free fatty acids salt band at 1570 cm-1. Although this new method did not provide a
substantial gain in terms of sensitivity (~ 0.2% FF A) over the standard and other
titrimetric methods, it offers both specificity to free fatty acids and analysis independent
of oil type, with differential spectroscopy being employed to circumvent matrix effects.
Moreover, unlike the titrimetric methods, thi~ FTIR method does not require reagent
standardization, electrode conditioning or maintenance.
The subsequent FF A method described in Chapter 4, takes a different approach to
the sarne analysis that makes it simpler to perforrn, more rapid and more sensitive without
losing specificity. The method is extraction-based and the target constituent (FF A in the
forrn of its salt) is extracted from oil sarnples using a polar oil-immiscible solvent
(methanol) containing a weak base (sodium hydrogen cyanamide) which is soluble in the
extracting solvent. Not only does this simplify the extraction procedure but it also
significantly enhances sensitivity and obviates oil matrix effects as the oil is no longer
present in the sarnple being analyzed spectroscopically. This method provides a simple
and robust means by which to analyze for low levels of FF A with a high degree of
104
accuracy (± 0.0006%) with low solvent consumption (2 ml/analysis). Furthermore, with a
simple adjustment of the reagent:oil ratio, the method is scalable for the analysis of semi
refined and crude oils containing up to 4.0% FFA. An additional benefit ofthis method is
that its sensitivity opens the door to more accurate monitoring of the refining processes
carried out on edible oils and the possible use of FF As as an indicator of lipid oxidation.
In Chapter 5, the sarne principle of extraction and concentration was used to
develop a rapid, practical, and accurate FTIR method for the determination of moisture
content in edible oils. This was possible because of the high polarity of water, its strong
mid IR absorption as weIl as the unique solvating characteristics of acetonitrile obviating
the need to carry out any conversion reactions to pro duce more IR-active species. For the
greatest sensitivity (0-200 ppm) the -OH stretching band (3629 cm- I) is used, while for
higher moisture levels (200-2000 ppm) the H-O-H bending band (1631 cm- I) of water is
suitable. One unique feature of the method is that the acetonitrile solvent does not need to
be absolutely dry to obtain accurate results, as the method corrects for traces of moi sture
present via spectral ratioing against the acetonitrile used to extract the sarnple, greatly
simplifying the methodology. It was also shown that a1cohols, hydroperoxides, and FF As,
which may be extracted into acetonitrile, do not interfere with the analysis. This robust
method has the potential to be a practical alternative to the Karl Fischer method, which
has many variables associated with its analytical performance.
The work in Chapter 6 concerns the implementation of the FF A and moi sture
extraction-based methods on an FTIR spectrometer coupled to an auto-sarnpler to
automate these analyses. The system developed effectively automates sample loading,
spectral collection and calculation of results using pre-prograrnmed algorithrns, with
output to screen or to a Laboratory Information Management System. It was tested,
optimized and validated, and shown to be capable of sarnple throughputs of >60
sarnpleslh with an accuracy very similar to that attained by using the manual procedures.
This system is designed for high-volurne contract, process or payment laboratories and
lays the groundwork for extending its utility to other common edible oil analyses, such as
peroxide value or trans content.
105
The development of the methods described in this thesis has been based on the
design of novel approaches to problems that have hindered the full exploitation of the
power of FTIR spectroscopy as a rapid and simple instrumental tool for edible oil
analysis. This research has provided very accurate and specific measurements for two of
the most important quality factors associated with edible oils. FF A and moisture methods
have been developed that are superior in specificity and simplicity to their respective
conventional titrimetric methods while providing comparable or better sensitivity and
accuracy. Furthermore, the integration of an FTIR spectrometer with an auto-sampler,
pump and appropriate software has made automated analysis of edible oils for FF A and
moi sture a realistic possibility. The adoption of such methods by industry can have a
significant impact on the monitoring of oil processing operations and product quality
control.
106
